freebsd-dev/sys/netinet/sctp_usrreq.c

5172 lines
142 KiB
C
Raw Normal View History

/*-
* Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2008-2011, by Randall Stewart. All rights reserved.
* Copyright (c) 2008-2011, by Michael Tuexen. 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.
*/
/* $KAME: sctp_usrreq.c,v 1.48 2005/03/07 23:26:08 itojun Exp $ */
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <netinet/sctp_os.h>
#include <sys/proc.h>
#include <netinet/sctp_pcb.h>
#include <netinet/sctp_header.h>
#include <netinet/sctp_var.h>
#if defined(INET6)
#endif
#include <netinet/sctp_sysctl.h>
#include <netinet/sctp_output.h>
#include <netinet/sctp_uio.h>
#include <netinet/sctp_asconf.h>
#include <netinet/sctputil.h>
#include <netinet/sctp_indata.h>
#include <netinet/sctp_timer.h>
#include <netinet/sctp_auth.h>
#include <netinet/sctp_bsd_addr.h>
#include <netinet/udp.h>
extern struct sctp_cc_functions sctp_cc_functions[];
extern struct sctp_ss_functions sctp_ss_functions[];
void
sctp_init(void)
{
u_long sb_max_adj;
/* Initialize and modify the sysctled variables */
sctp_init_sysctls();
if ((nmbclusters / 8) > SCTP_ASOC_MAX_CHUNKS_ON_QUEUE)
SCTP_BASE_SYSCTL(sctp_max_chunks_on_queue) = (nmbclusters / 8);
/*
* Allow a user to take no more than 1/2 the number of clusters or
* the SB_MAX whichever is smaller for the send window.
*/
sb_max_adj = (u_long)((u_quad_t) (SB_MAX) * MCLBYTES / (MSIZE + MCLBYTES));
SCTP_BASE_SYSCTL(sctp_sendspace) = min(sb_max_adj,
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
(((uint32_t) nmbclusters / 2) * SCTP_DEFAULT_MAXSEGMENT));
/*
* Now for the recv window, should we take the same amount? or
* should I do 1/2 the SB_MAX instead in the SB_MAX min above. For
* now I will just copy.
*/
SCTP_BASE_SYSCTL(sctp_recvspace) = SCTP_BASE_SYSCTL(sctp_sendspace);
SCTP_BASE_VAR(first_time) = 0;
SCTP_BASE_VAR(sctp_pcb_initialized) = 0;
sctp_pcb_init();
#if defined(SCTP_PACKET_LOGGING)
SCTP_BASE_VAR(packet_log_writers) = 0;
SCTP_BASE_VAR(packet_log_end) = 0;
bzero(&SCTP_BASE_VAR(packet_log_buffer), SCTP_PACKET_LOG_SIZE);
#endif
}
void
sctp_finish(void)
{
sctp_pcb_finish();
}
- Macroizes the V6ONLY flag check. - Added a short time wait (not used yet) constant - Corrected the type of the crc32c table (it was unsigned long and really is a uint32_t - Got rid of the user of MHeaders until they are truely needed by lower layers. - Fixed an initialization problem in the readq structure (ordering was off). - Found yet another collision bug when the random number generator returns two numbers on one side (during a collision) that are the same. Also added some tracking of cookies that will go away when we know that we have the last collision bug gone. - Fixed an init bug for book_size_scale, that was causing Early FR code to run when it should not. - Fixed a flight size tracking bug that was associated with Early FR but due to above bug also effected all FR's - Fixed it so Max Burst also will apply to Fast Retransmit. - Fixed a bug in the temporary logging code that allowed a static log array overflow - hashinit_flags is now used. - Two last mcopym's were converted to the macro sctp_m_copym that has always been used by all other places - macro sctp_m_copym was converted to upper case. - We now validate sinfo_flags on input (we did not before). - Fixed a bug that prevented a user from sending data and immediately shuting down with one send operation. - Moved to use hashdestroy instead of free() in our macros. - Fixed an init problem in our timed_wait vtag where we did not fully initialize our time-wait blocks. - Timer stops were re-positioned. - A pcb cleanup method was added, however this probably will not be used in BSD.. unless we make module loadable protocols - I think this fixes the mysterious timer bug.. it was a ordering of locks problem in the way we did timers. It now conforms to the timeout(9) manual (except for the _drain part, we had to do this a different way due to locks). - Fixed error return code so we get either CONNREUSED or CONNRESET depending on where one is in progression - Purged an unused clone macro. - Fixed a read erro code issue where we were NOT getting the proper error when the connection was reset. - Purged an unused clone macro. - Fixed a read erro code issue where we were NOT getting the proper error when the connection was reset. Approved by: gnn
2007-01-15 15:12:10 +00:00
void
sctp_pathmtu_adjustment(struct sctp_inpcb *inp,
struct sctp_tcb *stcb,
struct sctp_nets *net,
uint16_t nxtsz)
{
struct sctp_tmit_chunk *chk;
uint16_t overhead;
/* Adjust that too */
stcb->asoc.smallest_mtu = nxtsz;
/* now off to subtract IP_DF flag if needed */
overhead = IP_HDR_SIZE;
if (sctp_auth_is_required_chunk(SCTP_DATA, stcb->asoc.peer_auth_chunks)) {
overhead += sctp_get_auth_chunk_len(stcb->asoc.peer_hmac_id);
}
TAILQ_FOREACH(chk, &stcb->asoc.send_queue, sctp_next) {
if ((chk->send_size + overhead) > nxtsz) {
chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
}
}
TAILQ_FOREACH(chk, &stcb->asoc.sent_queue, sctp_next) {
if ((chk->send_size + overhead) > nxtsz) {
/*
* For this guy we also mark for immediate resend
* since we sent to big of chunk
*/
chk->flags |= CHUNK_FLAGS_FRAGMENT_OK;
if (chk->sent < SCTP_DATAGRAM_RESEND) {
sctp_flight_size_decrease(chk);
sctp_total_flight_decrease(stcb, chk);
}
if (chk->sent != SCTP_DATAGRAM_RESEND) {
sctp_ucount_incr(stcb->asoc.sent_queue_retran_cnt);
}
chk->sent = SCTP_DATAGRAM_RESEND;
chk->rec.data.doing_fast_retransmit = 0;
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_FLIGHT_LOGGING_ENABLE) {
sctp_misc_ints(SCTP_FLIGHT_LOG_DOWN_PMTU,
chk->whoTo->flight_size,
chk->book_size,
(uintptr_t) chk->whoTo,
chk->rec.data.TSN_seq);
}
/* Clear any time so NO RTT is being done */
chk->do_rtt = 0;
}
}
}
#ifdef INET
static void
sctp_notify_mbuf(struct sctp_inpcb *inp,
struct sctp_tcb *stcb,
struct sctp_nets *net,
struct ip *ip,
struct sctphdr *sh)
{
struct icmp *icmph;
int totsz, tmr_stopped = 0;
uint16_t nxtsz;
/* protection */
if ((inp == NULL) || (stcb == NULL) || (net == NULL) ||
(ip == NULL) || (sh == NULL)) {
if (stcb != NULL) {
SCTP_TCB_UNLOCK(stcb);
}
return;
}
/* First job is to verify the vtag matches what I would send */
if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag)) {
SCTP_TCB_UNLOCK(stcb);
return;
}
icmph = (struct icmp *)((caddr_t)ip - (sizeof(struct icmp) -
sizeof(struct ip)));
if (icmph->icmp_type != ICMP_UNREACH) {
/* We only care about unreachable */
SCTP_TCB_UNLOCK(stcb);
return;
}
if (icmph->icmp_code != ICMP_UNREACH_NEEDFRAG) {
/* not a unreachable message due to frag. */
SCTP_TCB_UNLOCK(stcb);
return;
}
totsz = ip->ip_len;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
nxtsz = ntohs(icmph->icmp_nextmtu);
if (nxtsz == 0) {
/*
* old type router that does not tell us what the next size
* mtu is. Rats we will have to guess (in a educated fashion
* of course)
*/
nxtsz = sctp_get_prev_mtu(totsz);
}
/* Stop any PMTU timer */
if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) {
tmr_stopped = 1;
sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net,
SCTP_FROM_SCTP_USRREQ + SCTP_LOC_1);
}
/* Adjust destination size limit */
if (net->mtu > nxtsz) {
net->mtu = nxtsz;
if (net->port) {
net->mtu -= sizeof(struct udphdr);
}
}
/* now what about the ep? */
if (stcb->asoc.smallest_mtu > nxtsz) {
sctp_pathmtu_adjustment(inp, stcb, net, nxtsz);
}
if (tmr_stopped)
sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
SCTP_TCB_UNLOCK(stcb);
}
#endif
void
sctp_notify(struct sctp_inpcb *inp,
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
struct ip *ip,
struct sctphdr *sh,
struct sockaddr *to,
struct sctp_tcb *stcb,
struct sctp_nets *net)
{
#if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
struct socket *so;
#endif
/* protection */
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
int reason;
struct icmp *icmph;
if ((inp == NULL) || (stcb == NULL) || (net == NULL) ||
(sh == NULL) || (to == NULL)) {
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
if (stcb)
SCTP_TCB_UNLOCK(stcb);
return;
}
/* First job is to verify the vtag matches what I would send */
if (ntohl(sh->v_tag) != (stcb->asoc.peer_vtag)) {
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
SCTP_TCB_UNLOCK(stcb);
return;
}
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
icmph = (struct icmp *)((caddr_t)ip - (sizeof(struct icmp) -
sizeof(struct ip)));
if (icmph->icmp_type != ICMP_UNREACH) {
/* We only care about unreachable */
SCTP_TCB_UNLOCK(stcb);
return;
}
if ((icmph->icmp_code == ICMP_UNREACH_NET) ||
(icmph->icmp_code == ICMP_UNREACH_HOST) ||
(icmph->icmp_code == ICMP_UNREACH_NET_UNKNOWN) ||
(icmph->icmp_code == ICMP_UNREACH_HOST_UNKNOWN) ||
(icmph->icmp_code == ICMP_UNREACH_ISOLATED) ||
(icmph->icmp_code == ICMP_UNREACH_NET_PROHIB) ||
(icmph->icmp_code == ICMP_UNREACH_HOST_PROHIB) ||
(icmph->icmp_code == ICMP_UNREACH_FILTER_PROHIB)) {
/*
* Hmm reachablity problems we must examine closely. If its
* not reachable, we may have lost a network. Or if there is
* NO protocol at the other end named SCTP. well we consider
* it a OOTB abort.
*/
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
if (net->dest_state & SCTP_ADDR_REACHABLE) {
/* Ok that destination is NOT reachable */
SCTP_PRINTF("ICMP (thresh %d/%d) takes interface %p down\n",
net->error_count,
net->failure_threshold,
net);
net->dest_state &= ~SCTP_ADDR_REACHABLE;
net->dest_state |= SCTP_ADDR_NOT_REACHABLE;
/*
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
* JRS 5/14/07 - If a destination is unreachable,
* the PF bit is turned off. This allows an
* unambiguous use of the PF bit for destinations
* that are reachable but potentially failed. If the
* destination is set to the unreachable state, also
* set the destination to the PF state.
*/
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
/*
* Add debug message here if destination is not in
* PF state.
*/
/* Stop any running T3 timers here? */
if ((stcb->asoc.sctp_cmt_on_off > 0) &&
(stcb->asoc.sctp_cmt_pf > 0)) {
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
net->dest_state &= ~SCTP_ADDR_PF;
SCTPDBG(SCTP_DEBUG_TIMER4, "Destination %p moved from PF to unreachable.\n",
net);
}
net->error_count = net->failure_threshold + 1;
sctp_ulp_notify(SCTP_NOTIFY_INTERFACE_DOWN,
stcb, SCTP_FAILED_THRESHOLD,
(void *)net, SCTP_SO_NOT_LOCKED);
}
SCTP_TCB_UNLOCK(stcb);
} else if ((icmph->icmp_code == ICMP_UNREACH_PROTOCOL) ||
(icmph->icmp_code == ICMP_UNREACH_PORT)) {
/*
* Here the peer is either playing tricks on us, including
* an address that belongs to someone who does not support
* SCTP OR was a userland implementation that shutdown and
* now is dead. In either case treat it like a OOTB abort
* with no TCB
*/
reason = SCTP_PEER_FAULTY;
sctp_abort_notification(stcb, reason, SCTP_SO_NOT_LOCKED);
#if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
so = SCTP_INP_SO(inp);
atomic_add_int(&stcb->asoc.refcnt, 1);
SCTP_TCB_UNLOCK(stcb);
SCTP_SOCKET_LOCK(so, 1);
SCTP_TCB_LOCK(stcb);
atomic_subtract_int(&stcb->asoc.refcnt, 1);
#endif
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
(void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_2);
#if defined (__APPLE__) || defined(SCTP_SO_LOCK_TESTING)
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
SCTP_SOCKET_UNLOCK(so, 1);
/* SCTP_TCB_UNLOCK(stcb); MT: I think this is not needed. */
#endif
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
/* no need to unlock here, since the TCB is gone */
} else {
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
SCTP_TCB_UNLOCK(stcb);
}
}
#ifdef INET
void
sctp_ctlinput(cmd, sa, vip)
int cmd;
struct sockaddr *sa;
void *vip;
{
struct ip *ip = vip;
struct sctphdr *sh;
uint32_t vrf_id;
/* FIX, for non-bsd is this right? */
vrf_id = SCTP_DEFAULT_VRFID;
if (sa->sa_family != AF_INET ||
((struct sockaddr_in *)sa)->sin_addr.s_addr == INADDR_ANY) {
return;
}
if (PRC_IS_REDIRECT(cmd)) {
ip = 0;
} else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) {
return;
}
if (ip) {
struct sctp_inpcb *inp = NULL;
struct sctp_tcb *stcb = NULL;
struct sctp_nets *net = NULL;
struct sockaddr_in to, from;
sh = (struct sctphdr *)((caddr_t)ip + (ip->ip_hl << 2));
bzero(&to, sizeof(to));
bzero(&from, sizeof(from));
from.sin_family = to.sin_family = AF_INET;
from.sin_len = to.sin_len = sizeof(to);
from.sin_port = sh->src_port;
from.sin_addr = ip->ip_src;
to.sin_port = sh->dest_port;
to.sin_addr = ip->ip_dst;
/*
* 'to' holds the dest of the packet that failed to be sent.
* 'from' holds our local endpoint address. Thus we reverse
* the to and the from in the lookup.
*/
stcb = sctp_findassociation_addr_sa((struct sockaddr *)&from,
(struct sockaddr *)&to,
&inp, &net, 1, vrf_id);
if (stcb != NULL && inp && (inp->sctp_socket != NULL)) {
if (cmd != PRC_MSGSIZE) {
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
sctp_notify(inp, ip, sh,
(struct sockaddr *)&to, stcb,
net);
} else {
/* handle possible ICMP size messages */
sctp_notify_mbuf(inp, stcb, net, ip, sh);
}
} else {
if ((stcb == NULL) && (inp != NULL)) {
/* reduce ref-count */
SCTP_INP_WLOCK(inp);
SCTP_INP_DECR_REF(inp);
SCTP_INP_WUNLOCK(inp);
}
if (stcb) {
SCTP_TCB_UNLOCK(stcb);
}
}
}
return;
}
#endif
static int
sctp_getcred(SYSCTL_HANDLER_ARGS)
{
struct xucred xuc;
struct sockaddr_in addrs[2];
struct sctp_inpcb *inp;
struct sctp_nets *net;
struct sctp_tcb *stcb;
int error;
uint32_t vrf_id;
/* FIX, for non-bsd is this right? */
vrf_id = SCTP_DEFAULT_VRFID;
error = priv_check(req->td, PRIV_NETINET_GETCRED);
if (error)
return (error);
error = SYSCTL_IN(req, addrs, sizeof(addrs));
if (error)
return (error);
stcb = sctp_findassociation_addr_sa(sintosa(&addrs[0]),
sintosa(&addrs[1]),
&inp, &net, 1, vrf_id);
if (stcb == NULL || inp == NULL || inp->sctp_socket == NULL) {
if ((inp != NULL) && (stcb == NULL)) {
/* reduce ref-count */
SCTP_INP_WLOCK(inp);
SCTP_INP_DECR_REF(inp);
goto cred_can_cont;
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
error = ENOENT;
goto out;
}
SCTP_TCB_UNLOCK(stcb);
/*
* We use the write lock here, only since in the error leg we need
* it. If we used RLOCK, then we would have to
* wlock/decr/unlock/rlock. Which in theory could create a hole.
* Better to use higher wlock.
*/
SCTP_INP_WLOCK(inp);
cred_can_cont:
error = cr_canseesocket(req->td->td_ucred, inp->sctp_socket);
if (error) {
SCTP_INP_WUNLOCK(inp);
goto out;
}
cru2x(inp->sctp_socket->so_cred, &xuc);
SCTP_INP_WUNLOCK(inp);
error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred));
out:
return (error);
}
SYSCTL_PROC(_net_inet_sctp, OID_AUTO, getcred, CTLTYPE_OPAQUE | CTLFLAG_RW,
0, 0, sctp_getcred, "S,ucred", "Get the ucred of a SCTP connection");
#ifdef INET
static void
sctp_abort(struct socket *so)
{
struct sctp_inpcb *inp;
uint32_t flags;
inp = (struct sctp_inpcb *)so->so_pcb;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (inp == 0) {
return;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
sctp_must_try_again:
flags = inp->sctp_flags;
#ifdef SCTP_LOG_CLOSING
sctp_log_closing(inp, NULL, 17);
#endif
if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) &&
(atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) {
#ifdef SCTP_LOG_CLOSING
sctp_log_closing(inp, NULL, 16);
#endif
sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT,
SCTP_CALLED_AFTER_CMPSET_OFCLOSE);
SOCK_LOCK(so);
SCTP_SB_CLEAR(so->so_snd);
/*
* same for the rcv ones, they are only here for the
* accounting/select.
*/
SCTP_SB_CLEAR(so->so_rcv);
/* Now null out the reference, we are completely detached. */
so->so_pcb = NULL;
SOCK_UNLOCK(so);
} else {
flags = inp->sctp_flags;
if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) {
goto sctp_must_try_again;
}
}
return;
}
static int
sctp_attach(struct socket *so, int proto, struct thread *p)
{
struct sctp_inpcb *inp;
struct inpcb *ip_inp;
int error;
uint32_t vrf_id = SCTP_DEFAULT_VRFID;
#ifdef IPSEC
uint32_t flags;
#endif
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp != 0) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return EINVAL;
}
if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
error = SCTP_SORESERVE(so, SCTP_BASE_SYSCTL(sctp_sendspace), SCTP_BASE_SYSCTL(sctp_recvspace));
if (error) {
return error;
}
}
error = sctp_inpcb_alloc(so, vrf_id);
if (error) {
return error;
}
inp = (struct sctp_inpcb *)so->so_pcb;
SCTP_INP_WLOCK(inp);
inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUND_V6; /* I'm not v6! */
ip_inp = &inp->ip_inp.inp;
ip_inp->inp_vflag |= INP_IPV4;
2009-09-17 15:11:12 +00:00
ip_inp->inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
#ifdef IPSEC
error = ipsec_init_policy(so, &ip_inp->inp_sp);
#ifdef SCTP_LOG_CLOSING
sctp_log_closing(inp, NULL, 17);
#endif
if (error != 0) {
try_again:
flags = inp->sctp_flags;
if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) &&
(atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) {
#ifdef SCTP_LOG_CLOSING
sctp_log_closing(inp, NULL, 15);
#endif
SCTP_INP_WUNLOCK(inp);
sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT,
SCTP_CALLED_AFTER_CMPSET_OFCLOSE);
} else {
flags = inp->sctp_flags;
if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) {
goto try_again;
} else {
SCTP_INP_WUNLOCK(inp);
}
}
return error;
}
#endif /* IPSEC */
SCTP_INP_WUNLOCK(inp);
return 0;
}
static int
sctp_bind(struct socket *so, struct sockaddr *addr, struct thread *p)
{
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
struct sctp_inpcb *inp = NULL;
int error;
#ifdef INET6
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (addr && addr->sa_family != AF_INET) {
/* must be a v4 address! */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
#endif /* INET6 */
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (addr && (addr->sa_len != sizeof(struct sockaddr_in))) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
return EINVAL;
}
inp = (struct sctp_inpcb *)so->so_pcb;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (inp == 0) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
- take out a needless panic under invariants for sctp_output.c - Fix addrs's error checking of sctp_sendx(3) when addrcnt is less than SCTP_SMALL_IOVEC_SIZE - re-add back inpcb_bind local address check bypass capability - Fix it so sctp_opt_info is independant of assoc_id postion. - Fix cookie life set to use MSEC_TO_TICKS() macro. - asconf changes o More comment changes/clarifications related to the old local address "not" list which is now an explicit restricted list. o Rename some functions for clarity: - sctp_add/del_local_addr_assoc to xxx_local_addr_restricted() - asconf related iterator functions to sctp_asconf_iterator_xxx() o Fix bug when the same address is deleted and added (and removed from the asconf queue) where the ifa is "freed" twice refcount wise, possibly freeing it completely. o Fix bug in output where the first ASCONF would not go out after the last address is changed (e.g. only goes out when retransmitted). o Fix bug where multiple ASCONFs can be bundled in the same packet with the and with the same serial numbers. o Fix asconf stcb iterator to not send ASCONF until after all work queue entries have been processed. o Change behavior so that when the last address is deleted (auto asconf on a bound all endpoint) no action is taken until an address is added; at that time, an ASCONF add+delete is sent (if the assoc is still up). o Fix local address counting so that address scoping is taken into account. o #ifdef SCTP_TIMER_BASED_ASCONF the old timer triggered sending of ASCONF (after an RTO). The default now is to send ASCONF immediately (except for the case of changing/deleting the last usable address). Approved by: re(ken smith)@freebsd.org
2007-07-24 20:06:02 +00:00
error = sctp_inpcb_bind(so, addr, NULL, p);
return error;
}
#endif
- During shutdown pending, when the last sack came in and the last message on the send stream was "null" but still there, a state we allow, we could get hung and not clean it up and wait for the shutdown guard timer to clear the association without a graceful close. Fix this so that that we properly clean up. - Added support for Multiple ASCONF per new RFC. We only (so far) accept input of these and cannot yet generate a multi-asconf. - Sysctl'd support for experimental Fast Handover feature. Always disabled unless sysctl or socket option changes to enable. - Error case in add-ip where the peer supports AUTH and ADD-IP but does NOT require AUTH of ASCONF/ASCONF-ACK. We need to ABORT in this case. - According to the Kyoto summit of socket api developers (Solaris, Linux, BSD). We need to have: o non-eeor mode messages be atomic - Fixed o Allow implicit setup of an assoc in 1-2-1 model if using the sctp_**() send calls - Fixed o Get rid of HAVE_XXX declarations - Done o add a sctp_pr_policy in hole in sndrcvinfo structure - Done o add a PR_SCTP_POLICY_VALID type flag - yet to-do in a future patch! - Optimize sctp6 calls to reuse code in sctp_usrreq. Also optimize when we close sending out the data and disabling Nagle. - Change key concatenation order to match the auth RFC - When sending OOTB shutdown_complete always do csum. - Don't send PKT-DROP to a PKT-DROP - For abort chunks just always checksums same for shutdown-complete. - inpcb_free front state had a bug where in queue data could wedge an assoc. We need to just abandon ones in front states (free_assoc). - If a peer sends us a 64k abort, we would try to assemble a response packet which may be larger than 64k. This then would be dropped by IP. Instead make a "minimum" size for us 64k-2k (we want at least 2k for our initack). If we receive such an init discard it early without all the processing. - When we peel off we must increment the tcb ref count to keep it from being freed from underneath us. - handling fwd-tsn had bugs that caused memory overwrites when given faulty data, fixed so can't happen and we also stop at the first bad stream no. - Fixed so comm-up generates the adaption indication. - peeloff did not get the hmac params copied. - fix it so we lock the addr list when doing src-addr selection (in future we need to use a multi-reader/one writer lock here) - During lowlevel output, we could end up with a _l_addr set to null if the iterator is calling the output routine. This means we would possibly crash when we gather the MTU info. Fix so we only do the gather where we have a src address cached. - we need to be sure to set abort flag on conn state when we receive an abort. - peeloff could leak a socket. Moved code so the close will find the socket if the peeloff fails (uipc_syscalls.c) Approved by: re@freebsd.org(Ken Smith)
2007-08-27 05:19:48 +00:00
void
sctp_close(struct socket *so)
{
struct sctp_inpcb *inp;
uint32_t flags;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0)
return;
/*
* Inform all the lower layer assoc that we are done.
*/
sctp_must_try_again:
flags = inp->sctp_flags;
#ifdef SCTP_LOG_CLOSING
sctp_log_closing(inp, NULL, 17);
#endif
if (((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) &&
(atomic_cmpset_int(&inp->sctp_flags, flags, (flags | SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_CLOSE_IP)))) {
if (((so->so_options & SO_LINGER) && (so->so_linger == 0)) ||
(so->so_rcv.sb_cc > 0)) {
#ifdef SCTP_LOG_CLOSING
sctp_log_closing(inp, NULL, 13);
#endif
sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_ABORT,
SCTP_CALLED_AFTER_CMPSET_OFCLOSE);
} else {
#ifdef SCTP_LOG_CLOSING
sctp_log_closing(inp, NULL, 14);
#endif
sctp_inpcb_free(inp, SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
SCTP_CALLED_AFTER_CMPSET_OFCLOSE);
}
/*
* The socket is now detached, no matter what the state of
* the SCTP association.
*/
SOCK_LOCK(so);
SCTP_SB_CLEAR(so->so_snd);
/*
* same for the rcv ones, they are only here for the
* accounting/select.
*/
SCTP_SB_CLEAR(so->so_rcv);
/* Now null out the reference, we are completely detached. */
so->so_pcb = NULL;
SOCK_UNLOCK(so);
} else {
flags = inp->sctp_flags;
if ((flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) {
goto sctp_must_try_again;
}
}
return;
}
int
sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
struct mbuf *control, struct thread *p);
int
sctp_sendm(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
struct mbuf *control, struct thread *p)
{
struct sctp_inpcb *inp;
int error;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0) {
if (control) {
sctp_m_freem(control);
control = NULL;
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
sctp_m_freem(m);
return EINVAL;
}
/* Got to have an to address if we are NOT a connected socket */
if ((addr == NULL) &&
((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) ||
(inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE))
) {
goto connected_type;
} else if (addr == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EDESTADDRREQ);
error = EDESTADDRREQ;
sctp_m_freem(m);
if (control) {
sctp_m_freem(control);
control = NULL;
}
return (error);
}
#ifdef INET6
if (addr->sa_family != AF_INET) {
/* must be a v4 address! */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET_PKT(m, inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EDESTADDRREQ);
sctp_m_freem(m);
if (control) {
sctp_m_freem(control);
control = NULL;
}
error = EDESTADDRREQ;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
return EDESTADDRREQ;
}
#endif /* INET6 */
connected_type:
/* now what about control */
if (control) {
if (inp->control) {
SCTP_PRINTF("huh? control set?\n");
sctp_m_freem(inp->control);
inp->control = NULL;
}
inp->control = control;
}
/* Place the data */
if (inp->pkt) {
SCTP_BUF_NEXT(inp->pkt_last) = m;
inp->pkt_last = m;
} else {
inp->pkt_last = inp->pkt = m;
}
if (
/* FreeBSD uses a flag passed */
((flags & PRUS_MORETOCOME) == 0)
) {
/*
* note with the current version this code will only be used
* by OpenBSD-- NetBSD, FreeBSD, and MacOS have methods for
* re-defining sosend to use the sctp_sosend. One can
* optionally switch back to this code (by changing back the
* definitions) but this is not advisable. This code is used
* by FreeBSD when sending a file with sendfile() though.
*/
int ret;
ret = sctp_output(inp, inp->pkt, addr, inp->control, p, flags);
inp->pkt = NULL;
inp->control = NULL;
return (ret);
} else {
return (0);
}
}
- During shutdown pending, when the last sack came in and the last message on the send stream was "null" but still there, a state we allow, we could get hung and not clean it up and wait for the shutdown guard timer to clear the association without a graceful close. Fix this so that that we properly clean up. - Added support for Multiple ASCONF per new RFC. We only (so far) accept input of these and cannot yet generate a multi-asconf. - Sysctl'd support for experimental Fast Handover feature. Always disabled unless sysctl or socket option changes to enable. - Error case in add-ip where the peer supports AUTH and ADD-IP but does NOT require AUTH of ASCONF/ASCONF-ACK. We need to ABORT in this case. - According to the Kyoto summit of socket api developers (Solaris, Linux, BSD). We need to have: o non-eeor mode messages be atomic - Fixed o Allow implicit setup of an assoc in 1-2-1 model if using the sctp_**() send calls - Fixed o Get rid of HAVE_XXX declarations - Done o add a sctp_pr_policy in hole in sndrcvinfo structure - Done o add a PR_SCTP_POLICY_VALID type flag - yet to-do in a future patch! - Optimize sctp6 calls to reuse code in sctp_usrreq. Also optimize when we close sending out the data and disabling Nagle. - Change key concatenation order to match the auth RFC - When sending OOTB shutdown_complete always do csum. - Don't send PKT-DROP to a PKT-DROP - For abort chunks just always checksums same for shutdown-complete. - inpcb_free front state had a bug where in queue data could wedge an assoc. We need to just abandon ones in front states (free_assoc). - If a peer sends us a 64k abort, we would try to assemble a response packet which may be larger than 64k. This then would be dropped by IP. Instead make a "minimum" size for us 64k-2k (we want at least 2k for our initack). If we receive such an init discard it early without all the processing. - When we peel off we must increment the tcb ref count to keep it from being freed from underneath us. - handling fwd-tsn had bugs that caused memory overwrites when given faulty data, fixed so can't happen and we also stop at the first bad stream no. - Fixed so comm-up generates the adaption indication. - peeloff did not get the hmac params copied. - fix it so we lock the addr list when doing src-addr selection (in future we need to use a multi-reader/one writer lock here) - During lowlevel output, we could end up with a _l_addr set to null if the iterator is calling the output routine. This means we would possibly crash when we gather the MTU info. Fix so we only do the gather where we have a src address cached. - we need to be sure to set abort flag on conn state when we receive an abort. - peeloff could leak a socket. Moved code so the close will find the socket if the peeloff fails (uipc_syscalls.c) Approved by: re@freebsd.org(Ken Smith)
2007-08-27 05:19:48 +00:00
int
sctp_disconnect(struct socket *so)
{
struct sctp_inpcb *inp;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN);
return (ENOTCONN);
}
SCTP_INP_RLOCK(inp);
if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
if (LIST_EMPTY(&inp->sctp_asoc_list)) {
/* No connection */
SCTP_INP_RUNLOCK(inp);
return (0);
} else {
struct sctp_association *asoc;
struct sctp_tcb *stcb;
stcb = LIST_FIRST(&inp->sctp_asoc_list);
if (stcb == NULL) {
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (EINVAL);
}
SCTP_TCB_LOCK(stcb);
asoc = &stcb->asoc;
if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
/* We are about to be freed, out of here */
SCTP_TCB_UNLOCK(stcb);
SCTP_INP_RUNLOCK(inp);
return (0);
}
if (((so->so_options & SO_LINGER) &&
(so->so_linger == 0)) ||
(so->so_rcv.sb_cc > 0)) {
if (SCTP_GET_STATE(asoc) !=
SCTP_STATE_COOKIE_WAIT) {
/* Left with Data unread */
struct mbuf *err;
err = sctp_get_mbuf_for_msg(sizeof(struct sctp_paramhdr), 0, M_DONTWAIT, 1, MT_DATA);
if (err) {
/*
* Fill in the user
* initiated abort
*/
struct sctp_paramhdr *ph;
ph = mtod(err, struct sctp_paramhdr *);
SCTP_BUF_LEN(err) = sizeof(struct sctp_paramhdr);
ph->param_type = htons(SCTP_CAUSE_USER_INITIATED_ABT);
ph->param_length = htons(SCTP_BUF_LEN(err));
}
#if defined(SCTP_PANIC_ON_ABORT)
panic("disconnect does an abort");
#endif
sctp_send_abort_tcb(stcb, err, SCTP_SO_LOCKED);
SCTP_STAT_INCR_COUNTER32(sctps_aborted);
}
SCTP_INP_RUNLOCK(inp);
if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) ||
(SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
SCTP_STAT_DECR_GAUGE32(sctps_currestab);
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
(void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_3);
/* No unlock tcb assoc is gone */
return (0);
}
if (TAILQ_EMPTY(&asoc->send_queue) &&
TAILQ_EMPTY(&asoc->sent_queue) &&
(asoc->stream_queue_cnt == 0)) {
/* there is nothing queued to send, so done */
if (asoc->locked_on_sending) {
goto abort_anyway;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if ((SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
(SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
/* only send SHUTDOWN 1st time thru */
sctp_stop_timers_for_shutdown(stcb);
sctp_send_shutdown(stcb,
stcb->asoc.primary_destination);
sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_LOCKED);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
(SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
SCTP_STAT_DECR_GAUGE32(sctps_currestab);
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
stcb->sctp_ep, stcb,
asoc->primary_destination);
sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
stcb->sctp_ep, stcb,
asoc->primary_destination);
}
} else {
/*
* we still got (or just got) data to send,
* so set SHUTDOWN_PENDING
*/
/*
* XXX sockets draft says that SCTP_EOF
* should be sent with no data. currently,
* we will allow user data to be sent first
* and move to SHUTDOWN-PENDING
*/
asoc->state |= SCTP_STATE_SHUTDOWN_PENDING;
sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb,
asoc->primary_destination);
if (asoc->locked_on_sending) {
/* Locked to send out the data */
struct sctp_stream_queue_pending *sp;
sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead);
if (sp == NULL) {
SCTP_PRINTF("Error, sp is NULL, locked on sending is non-null strm:%d\n",
asoc->locked_on_sending->stream_no);
} else {
if ((sp->length == 0) && (sp->msg_is_complete == 0))
asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT;
}
}
if (TAILQ_EMPTY(&asoc->send_queue) &&
TAILQ_EMPTY(&asoc->sent_queue) &&
(asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
struct mbuf *op_err;
abort_anyway:
op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
0, M_DONTWAIT, 1, MT_DATA);
if (op_err) {
/*
* Fill in the user
* initiated abort
*/
struct sctp_paramhdr *ph;
uint32_t *ippp;
SCTP_BUF_LEN(op_err) =
(sizeof(struct sctp_paramhdr) + sizeof(uint32_t));
ph = mtod(op_err,
struct sctp_paramhdr *);
ph->param_type = htons(
SCTP_CAUSE_USER_INITIATED_ABT);
ph->param_length = htons(SCTP_BUF_LEN(op_err));
ippp = (uint32_t *) (ph + 1);
*ippp = htonl(SCTP_FROM_SCTP_USRREQ + SCTP_LOC_4);
}
#if defined(SCTP_PANIC_ON_ABORT)
panic("disconnect does an abort");
#endif
stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_4;
sctp_send_abort_tcb(stcb, op_err, SCTP_SO_LOCKED);
SCTP_STAT_INCR_COUNTER32(sctps_aborted);
if ((SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_OPEN) ||
(SCTP_GET_STATE(&stcb->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
SCTP_STAT_DECR_GAUGE32(sctps_currestab);
}
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
(void)sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_5);
return (0);
- During shutdown pending, when the last sack came in and the last message on the send stream was "null" but still there, a state we allow, we could get hung and not clean it up and wait for the shutdown guard timer to clear the association without a graceful close. Fix this so that that we properly clean up. - Added support for Multiple ASCONF per new RFC. We only (so far) accept input of these and cannot yet generate a multi-asconf. - Sysctl'd support for experimental Fast Handover feature. Always disabled unless sysctl or socket option changes to enable. - Error case in add-ip where the peer supports AUTH and ADD-IP but does NOT require AUTH of ASCONF/ASCONF-ACK. We need to ABORT in this case. - According to the Kyoto summit of socket api developers (Solaris, Linux, BSD). We need to have: o non-eeor mode messages be atomic - Fixed o Allow implicit setup of an assoc in 1-2-1 model if using the sctp_**() send calls - Fixed o Get rid of HAVE_XXX declarations - Done o add a sctp_pr_policy in hole in sndrcvinfo structure - Done o add a PR_SCTP_POLICY_VALID type flag - yet to-do in a future patch! - Optimize sctp6 calls to reuse code in sctp_usrreq. Also optimize when we close sending out the data and disabling Nagle. - Change key concatenation order to match the auth RFC - When sending OOTB shutdown_complete always do csum. - Don't send PKT-DROP to a PKT-DROP - For abort chunks just always checksums same for shutdown-complete. - inpcb_free front state had a bug where in queue data could wedge an assoc. We need to just abandon ones in front states (free_assoc). - If a peer sends us a 64k abort, we would try to assemble a response packet which may be larger than 64k. This then would be dropped by IP. Instead make a "minimum" size for us 64k-2k (we want at least 2k for our initack). If we receive such an init discard it early without all the processing. - When we peel off we must increment the tcb ref count to keep it from being freed from underneath us. - handling fwd-tsn had bugs that caused memory overwrites when given faulty data, fixed so can't happen and we also stop at the first bad stream no. - Fixed so comm-up generates the adaption indication. - peeloff did not get the hmac params copied. - fix it so we lock the addr list when doing src-addr selection (in future we need to use a multi-reader/one writer lock here) - During lowlevel output, we could end up with a _l_addr set to null if the iterator is calling the output routine. This means we would possibly crash when we gather the MTU info. Fix so we only do the gather where we have a src address cached. - we need to be sure to set abort flag on conn state when we receive an abort. - peeloff could leak a socket. Moved code so the close will find the socket if the peeloff fails (uipc_syscalls.c) Approved by: re@freebsd.org(Ken Smith)
2007-08-27 05:19:48 +00:00
} else {
sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
}
}
soisdisconnecting(so);
SCTP_TCB_UNLOCK(stcb);
SCTP_INP_RUNLOCK(inp);
return (0);
}
/* not reached */
} else {
/* UDP model does not support this */
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
return EOPNOTSUPP;
}
}
int
sctp_flush(struct socket *so, int how)
{
/*
* We will just clear out the values and let subsequent close clear
* out the data, if any. Note if the user did a shutdown(SHUT_RD)
* they will not be able to read the data, the socket will block
* that from happening.
*/
struct sctp_inpcb *inp;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == NULL) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return EINVAL;
}
SCTP_INP_RLOCK(inp);
/* For the 1 to many model this does nothing */
if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) {
SCTP_INP_RUNLOCK(inp);
return (0);
}
SCTP_INP_RUNLOCK(inp);
if ((how == PRU_FLUSH_RD) || (how == PRU_FLUSH_RDWR)) {
/*
* First make sure the sb will be happy, we don't use these
* except maybe the count
*/
SCTP_INP_WLOCK(inp);
SCTP_INP_READ_LOCK(inp);
inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_CANT_READ;
SCTP_INP_READ_UNLOCK(inp);
SCTP_INP_WUNLOCK(inp);
so->so_rcv.sb_cc = 0;
so->so_rcv.sb_mbcnt = 0;
so->so_rcv.sb_mb = NULL;
}
if ((how == PRU_FLUSH_WR) || (how == PRU_FLUSH_RDWR)) {
/*
* First make sure the sb will be happy, we don't use these
* except maybe the count
*/
so->so_snd.sb_cc = 0;
so->so_snd.sb_mbcnt = 0;
so->so_snd.sb_mb = NULL;
}
return (0);
}
int
sctp_shutdown(struct socket *so)
{
struct sctp_inpcb *inp;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return EINVAL;
}
SCTP_INP_RLOCK(inp);
/* For UDP model this is a invalid call */
if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) {
/* Restore the flags that the soshutdown took away. */
SOCKBUF_LOCK(&so->so_rcv);
so->so_rcv.sb_state &= ~SBS_CANTRCVMORE;
SOCKBUF_UNLOCK(&so->so_rcv);
/* This proc will wakeup for read and do nothing (I hope) */
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
return (EOPNOTSUPP);
}
/*
* Ok if we reach here its the TCP model and it is either a SHUT_WR
* or SHUT_RDWR. This means we put the shutdown flag against it.
*/
{
struct sctp_tcb *stcb;
struct sctp_association *asoc;
if ((so->so_state &
(SS_ISCONNECTED | SS_ISCONNECTING | SS_ISDISCONNECTING)) == 0) {
SCTP_INP_RUNLOCK(inp);
return (ENOTCONN);
}
socantsendmore(so);
stcb = LIST_FIRST(&inp->sctp_asoc_list);
if (stcb == NULL) {
/*
* Ok we hit the case that the shutdown call was
* made after an abort or something. Nothing to do
* now.
*/
SCTP_INP_RUNLOCK(inp);
return (0);
}
SCTP_TCB_LOCK(stcb);
asoc = &stcb->asoc;
if (TAILQ_EMPTY(&asoc->send_queue) &&
TAILQ_EMPTY(&asoc->sent_queue) &&
(asoc->stream_queue_cnt == 0)) {
if (asoc->locked_on_sending) {
goto abort_anyway;
}
/* there is nothing queued to send, so I'm done... */
if (SCTP_GET_STATE(asoc) != SCTP_STATE_SHUTDOWN_SENT) {
/* only send SHUTDOWN the first time through */
sctp_stop_timers_for_shutdown(stcb);
sctp_send_shutdown(stcb,
stcb->asoc.primary_destination);
sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_T3, SCTP_SO_LOCKED);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if ((SCTP_GET_STATE(asoc) == SCTP_STATE_OPEN) ||
(SCTP_GET_STATE(asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
SCTP_STAT_DECR_GAUGE32(sctps_currestab);
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_SET_STATE(asoc, SCTP_STATE_SHUTDOWN_SENT);
SCTP_CLEAR_SUBSTATE(asoc, SCTP_STATE_SHUTDOWN_PENDING);
sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN,
stcb->sctp_ep, stcb,
asoc->primary_destination);
sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD,
stcb->sctp_ep, stcb,
asoc->primary_destination);
}
} else {
/*
* we still got (or just got) data to send, so set
* SHUTDOWN_PENDING
*/
asoc->state |= SCTP_STATE_SHUTDOWN_PENDING;
sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, stcb->sctp_ep, stcb,
asoc->primary_destination);
if (asoc->locked_on_sending) {
/* Locked to send out the data */
struct sctp_stream_queue_pending *sp;
sp = TAILQ_LAST(&asoc->locked_on_sending->outqueue, sctp_streamhead);
if (sp == NULL) {
SCTP_PRINTF("Error, sp is NULL, locked on sending is non-null strm:%d\n",
asoc->locked_on_sending->stream_no);
} else {
if ((sp->length == 0) && (sp->msg_is_complete == 0)) {
asoc->state |= SCTP_STATE_PARTIAL_MSG_LEFT;
}
}
}
if (TAILQ_EMPTY(&asoc->send_queue) &&
TAILQ_EMPTY(&asoc->sent_queue) &&
(asoc->state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
struct mbuf *op_err;
abort_anyway:
op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
0, M_DONTWAIT, 1, MT_DATA);
if (op_err) {
/* Fill in the user initiated abort */
struct sctp_paramhdr *ph;
uint32_t *ippp;
SCTP_BUF_LEN(op_err) =
sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
ph = mtod(op_err,
struct sctp_paramhdr *);
ph->param_type = htons(
SCTP_CAUSE_USER_INITIATED_ABT);
ph->param_length = htons(SCTP_BUF_LEN(op_err));
ippp = (uint32_t *) (ph + 1);
*ippp = htonl(SCTP_FROM_SCTP_USRREQ + SCTP_LOC_6);
}
#if defined(SCTP_PANIC_ON_ABORT)
panic("shutdown does an abort");
#endif
stcb->sctp_ep->last_abort_code = SCTP_FROM_SCTP_USRREQ + SCTP_LOC_6;
sctp_abort_an_association(stcb->sctp_ep, stcb,
SCTP_RESPONSE_TO_USER_REQ,
op_err, SCTP_SO_LOCKED);
goto skip_unlock;
- During shutdown pending, when the last sack came in and the last message on the send stream was "null" but still there, a state we allow, we could get hung and not clean it up and wait for the shutdown guard timer to clear the association without a graceful close. Fix this so that that we properly clean up. - Added support for Multiple ASCONF per new RFC. We only (so far) accept input of these and cannot yet generate a multi-asconf. - Sysctl'd support for experimental Fast Handover feature. Always disabled unless sysctl or socket option changes to enable. - Error case in add-ip where the peer supports AUTH and ADD-IP but does NOT require AUTH of ASCONF/ASCONF-ACK. We need to ABORT in this case. - According to the Kyoto summit of socket api developers (Solaris, Linux, BSD). We need to have: o non-eeor mode messages be atomic - Fixed o Allow implicit setup of an assoc in 1-2-1 model if using the sctp_**() send calls - Fixed o Get rid of HAVE_XXX declarations - Done o add a sctp_pr_policy in hole in sndrcvinfo structure - Done o add a PR_SCTP_POLICY_VALID type flag - yet to-do in a future patch! - Optimize sctp6 calls to reuse code in sctp_usrreq. Also optimize when we close sending out the data and disabling Nagle. - Change key concatenation order to match the auth RFC - When sending OOTB shutdown_complete always do csum. - Don't send PKT-DROP to a PKT-DROP - For abort chunks just always checksums same for shutdown-complete. - inpcb_free front state had a bug where in queue data could wedge an assoc. We need to just abandon ones in front states (free_assoc). - If a peer sends us a 64k abort, we would try to assemble a response packet which may be larger than 64k. This then would be dropped by IP. Instead make a "minimum" size for us 64k-2k (we want at least 2k for our initack). If we receive such an init discard it early without all the processing. - When we peel off we must increment the tcb ref count to keep it from being freed from underneath us. - handling fwd-tsn had bugs that caused memory overwrites when given faulty data, fixed so can't happen and we also stop at the first bad stream no. - Fixed so comm-up generates the adaption indication. - peeloff did not get the hmac params copied. - fix it so we lock the addr list when doing src-addr selection (in future we need to use a multi-reader/one writer lock here) - During lowlevel output, we could end up with a _l_addr set to null if the iterator is calling the output routine. This means we would possibly crash when we gather the MTU info. Fix so we only do the gather where we have a src address cached. - we need to be sure to set abort flag on conn state when we receive an abort. - peeloff could leak a socket. Moved code so the close will find the socket if the peeloff fails (uipc_syscalls.c) Approved by: re@freebsd.org(Ken Smith)
2007-08-27 05:19:48 +00:00
} else {
sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
}
}
SCTP_TCB_UNLOCK(stcb);
}
skip_unlock:
SCTP_INP_RUNLOCK(inp);
return 0;
}
/*
* copies a "user" presentable address and removes embedded scope, etc.
* returns 0 on success, 1 on error
*/
static uint32_t
sctp_fill_user_address(struct sockaddr_storage *ss, struct sockaddr *sa)
{
#ifdef INET6
struct sockaddr_in6 lsa6;
sa = (struct sockaddr *)sctp_recover_scope((struct sockaddr_in6 *)sa,
&lsa6);
#endif
memcpy(ss, sa, sa->sa_len);
return (0);
}
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
/*
* NOTE: assumes addr lock is held
*/
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
static size_t
sctp_fill_up_addresses_vrf(struct sctp_inpcb *inp,
struct sctp_tcb *stcb,
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
size_t limit,
struct sockaddr_storage *sas,
uint32_t vrf_id)
{
struct sctp_ifn *sctp_ifn;
struct sctp_ifa *sctp_ifa;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
int loopback_scope, ipv4_local_scope, local_scope, site_scope;
size_t actual;
int ipv4_addr_legal, ipv6_addr_legal;
struct sctp_vrf *vrf;
actual = 0;
if (limit <= 0)
return (actual);
if (stcb) {
/* Turn on all the appropriate scope */
loopback_scope = stcb->asoc.loopback_scope;
ipv4_local_scope = stcb->asoc.ipv4_local_scope;
local_scope = stcb->asoc.local_scope;
site_scope = stcb->asoc.site_scope;
} else {
/* Turn on ALL scope, since we look at the EP */
loopback_scope = ipv4_local_scope = local_scope =
site_scope = 1;
}
ipv4_addr_legal = ipv6_addr_legal = 0;
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
ipv6_addr_legal = 1;
- Macroizes the V6ONLY flag check. - Added a short time wait (not used yet) constant - Corrected the type of the crc32c table (it was unsigned long and really is a uint32_t - Got rid of the user of MHeaders until they are truely needed by lower layers. - Fixed an initialization problem in the readq structure (ordering was off). - Found yet another collision bug when the random number generator returns two numbers on one side (during a collision) that are the same. Also added some tracking of cookies that will go away when we know that we have the last collision bug gone. - Fixed an init bug for book_size_scale, that was causing Early FR code to run when it should not. - Fixed a flight size tracking bug that was associated with Early FR but due to above bug also effected all FR's - Fixed it so Max Burst also will apply to Fast Retransmit. - Fixed a bug in the temporary logging code that allowed a static log array overflow - hashinit_flags is now used. - Two last mcopym's were converted to the macro sctp_m_copym that has always been used by all other places - macro sctp_m_copym was converted to upper case. - We now validate sinfo_flags on input (we did not before). - Fixed a bug that prevented a user from sending data and immediately shuting down with one send operation. - Moved to use hashdestroy instead of free() in our macros. - Fixed an init problem in our timed_wait vtag where we did not fully initialize our time-wait blocks. - Timer stops were re-positioned. - A pcb cleanup method was added, however this probably will not be used in BSD.. unless we make module loadable protocols - I think this fixes the mysterious timer bug.. it was a ordering of locks problem in the way we did timers. It now conforms to the timeout(9) manual (except for the _drain part, we had to do this a different way due to locks). - Fixed error return code so we get either CONNREUSED or CONNRESET depending on where one is in progression - Purged an unused clone macro. - Fixed a read erro code issue where we were NOT getting the proper error when the connection was reset. - Purged an unused clone macro. - Fixed a read erro code issue where we were NOT getting the proper error when the connection was reset. Approved by: gnn
2007-01-15 15:12:10 +00:00
if (SCTP_IPV6_V6ONLY(inp) == 0) {
ipv4_addr_legal = 1;
}
} else {
ipv4_addr_legal = 1;
}
vrf = sctp_find_vrf(vrf_id);
if (vrf == NULL) {
return (0);
}
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
if ((loopback_scope == 0) &&
SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
/* Skip loopback if loopback_scope not set */
continue;
}
LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
if (stcb) {
/*
* For the BOUND-ALL case, the list
* associated with a TCB is Always
* considered a reverse list.. i.e.
* it lists addresses that are NOT
* part of the association. If this
* is one of those we must skip it.
*/
if (sctp_is_addr_restricted(stcb,
sctp_ifa)) {
continue;
}
}
switch (sctp_ifa->address.sa.sa_family) {
#ifdef INET
case AF_INET:
if (ipv4_addr_legal) {
struct sockaddr_in *sin;
sin = (struct sockaddr_in *)&sctp_ifa->address.sa;
if (sin->sin_addr.s_addr == 0) {
/*
* we skip
* unspecifed
* addresses
*/
continue;
}
if ((ipv4_local_scope == 0) &&
(IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
continue;
}
#ifdef INET6
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) {
in6_sin_2_v4mapsin6(sin, (struct sockaddr_in6 *)sas);
((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport;
sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(struct sockaddr_in6));
actual += sizeof(struct sockaddr_in6);
} else {
#endif
memcpy(sas, sin, sizeof(*sin));
((struct sockaddr_in *)sas)->sin_port = inp->sctp_lport;
sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(*sin));
actual += sizeof(*sin);
#ifdef INET6
}
#endif
if (actual >= limit) {
return (actual);
}
} else {
continue;
}
break;
#endif
#ifdef INET6
case AF_INET6:
if (ipv6_addr_legal) {
struct sockaddr_in6 *sin6;
sin6 = (struct sockaddr_in6 *)&sctp_ifa->address.sa;
if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
/*
* we skip
* unspecifed
* addresses
*/
continue;
}
if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
if (local_scope == 0)
continue;
if (sin6->sin6_scope_id == 0) {
if (sa6_recoverscope(sin6) != 0)
/*
*
* bad
*
* li
* nk
*
* loc
* al
*
* add
* re
* ss
* */
continue;
}
}
if ((site_scope == 0) &&
(IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
continue;
}
memcpy(sas, sin6, sizeof(*sin6));
((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport;
sas = (struct sockaddr_storage *)((caddr_t)sas + sizeof(*sin6));
actual += sizeof(*sin6);
if (actual >= limit) {
return (actual);
}
} else {
continue;
}
break;
#endif
default:
/* TSNH */
break;
}
}
}
} else {
struct sctp_laddr *laddr;
LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
if (stcb) {
if (sctp_is_addr_restricted(stcb, laddr->ifa)) {
continue;
}
}
if (sctp_fill_user_address(sas, &laddr->ifa->address.sa))
continue;
((struct sockaddr_in6 *)sas)->sin6_port = inp->sctp_lport;
sas = (struct sockaddr_storage *)((caddr_t)sas +
laddr->ifa->address.sa.sa_len);
actual += laddr->ifa->address.sa.sa_len;
if (actual >= limit) {
return (actual);
}
}
}
return (actual);
}
static size_t
sctp_fill_up_addresses(struct sctp_inpcb *inp,
struct sctp_tcb *stcb,
size_t limit,
struct sockaddr_storage *sas)
{
size_t size = 0;
SCTP_IPI_ADDR_RLOCK();
/* fill up addresses for the endpoint's default vrf */
size = sctp_fill_up_addresses_vrf(inp, stcb, limit, sas,
inp->def_vrf_id);
SCTP_IPI_ADDR_RUNLOCK();
return (size);
}
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
/*
* NOTE: assumes addr lock is held
*/
static int
sctp_count_max_addresses_vrf(struct sctp_inpcb *inp, uint32_t vrf_id)
{
int cnt = 0;
struct sctp_vrf *vrf = NULL;
/*
* In both sub-set bound an bound_all cases we return the MAXIMUM
* number of addresses that you COULD get. In reality the sub-set
* bound may have an exclusion list for a given TCB OR in the
* bound-all case a TCB may NOT include the loopback or other
* addresses as well.
*/
vrf = sctp_find_vrf(vrf_id);
if (vrf == NULL) {
return (0);
}
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
struct sctp_ifn *sctp_ifn;
struct sctp_ifa *sctp_ifa;
LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
/* Count them if they are the right type */
switch (sctp_ifa->address.sa.sa_family) {
#ifdef INET
case AF_INET:
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4))
cnt += sizeof(struct sockaddr_in6);
else
cnt += sizeof(struct sockaddr_in);
break;
#endif
#ifdef INET6
case AF_INET6:
cnt += sizeof(struct sockaddr_in6);
break;
#endif
default:
break;
}
}
}
} else {
struct sctp_laddr *laddr;
LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
switch (laddr->ifa->address.sa.sa_family) {
#ifdef INET
case AF_INET:
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4))
cnt += sizeof(struct sockaddr_in6);
else
cnt += sizeof(struct sockaddr_in);
break;
#endif
#ifdef INET6
case AF_INET6:
cnt += sizeof(struct sockaddr_in6);
break;
#endif
default:
break;
}
}
}
return (cnt);
}
static int
sctp_count_max_addresses(struct sctp_inpcb *inp)
{
int cnt = 0;
SCTP_IPI_ADDR_RLOCK();
/* count addresses for the endpoint's default VRF */
cnt = sctp_count_max_addresses_vrf(inp, inp->def_vrf_id);
SCTP_IPI_ADDR_RUNLOCK();
return (cnt);
}
static int
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
sctp_do_connect_x(struct socket *so, struct sctp_inpcb *inp, void *optval,
size_t optsize, void *p, int delay)
{
int error = 0;
int creat_lock_on = 0;
struct sctp_tcb *stcb = NULL;
struct sockaddr *sa;
int num_v6 = 0, num_v4 = 0, *totaddrp, totaddr;
int added = 0;
uint32_t vrf_id;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
int bad_addresses = 0;
sctp_assoc_t *a_id;
SCTPDBG(SCTP_DEBUG_PCB1, "Connectx called\n");
if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) &&
(inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) {
/* We are already connected AND the TCP model */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE);
return (EADDRINUSE);
}
if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
(sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (EINVAL);
}
if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
SCTP_INP_RLOCK(inp);
stcb = LIST_FIRST(&inp->sctp_asoc_list);
SCTP_INP_RUNLOCK(inp);
}
if (stcb) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY);
return (EALREADY);
}
SCTP_INP_INCR_REF(inp);
SCTP_ASOC_CREATE_LOCK(inp);
creat_lock_on = 1;
if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
(inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT);
error = EFAULT;
goto out_now;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
totaddrp = (int *)optval;
totaddr = *totaddrp;
sa = (struct sockaddr *)(totaddrp + 1);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
stcb = sctp_connectx_helper_find(inp, sa, &totaddr, &num_v4, &num_v6, &error, (optsize - sizeof(int)), &bad_addresses);
if ((stcb != NULL) || bad_addresses) {
/* Already have or am bring up an association */
SCTP_ASOC_CREATE_UNLOCK(inp);
creat_lock_on = 0;
if (stcb)
SCTP_TCB_UNLOCK(stcb);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (bad_addresses == 0) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
error = EALREADY;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
goto out_now;
}
#ifdef INET6
if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) &&
(num_v6 > 0)) {
error = EINVAL;
goto out_now;
}
if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
(num_v4 > 0)) {
struct in6pcb *inp6;
inp6 = (struct in6pcb *)inp;
- Macroizes the V6ONLY flag check. - Added a short time wait (not used yet) constant - Corrected the type of the crc32c table (it was unsigned long and really is a uint32_t - Got rid of the user of MHeaders until they are truely needed by lower layers. - Fixed an initialization problem in the readq structure (ordering was off). - Found yet another collision bug when the random number generator returns two numbers on one side (during a collision) that are the same. Also added some tracking of cookies that will go away when we know that we have the last collision bug gone. - Fixed an init bug for book_size_scale, that was causing Early FR code to run when it should not. - Fixed a flight size tracking bug that was associated with Early FR but due to above bug also effected all FR's - Fixed it so Max Burst also will apply to Fast Retransmit. - Fixed a bug in the temporary logging code that allowed a static log array overflow - hashinit_flags is now used. - Two last mcopym's were converted to the macro sctp_m_copym that has always been used by all other places - macro sctp_m_copym was converted to upper case. - We now validate sinfo_flags on input (we did not before). - Fixed a bug that prevented a user from sending data and immediately shuting down with one send operation. - Moved to use hashdestroy instead of free() in our macros. - Fixed an init problem in our timed_wait vtag where we did not fully initialize our time-wait blocks. - Timer stops were re-positioned. - A pcb cleanup method was added, however this probably will not be used in BSD.. unless we make module loadable protocols - I think this fixes the mysterious timer bug.. it was a ordering of locks problem in the way we did timers. It now conforms to the timeout(9) manual (except for the _drain part, we had to do this a different way due to locks). - Fixed error return code so we get either CONNREUSED or CONNRESET depending on where one is in progression - Purged an unused clone macro. - Fixed a read erro code issue where we were NOT getting the proper error when the connection was reset. - Purged an unused clone macro. - Fixed a read erro code issue where we were NOT getting the proper error when the connection was reset. Approved by: gnn
2007-01-15 15:12:10 +00:00
if (SCTP_IPV6_V6ONLY(inp6)) {
/*
* if IPV6_V6ONLY flag, ignore connections destined
* to a v4 addr or v4-mapped addr
*/
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
goto out_now;
}
}
#endif /* INET6 */
if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) ==
SCTP_PCB_FLAGS_UNBOUND) {
/* Bind a ephemeral port */
- take out a needless panic under invariants for sctp_output.c - Fix addrs's error checking of sctp_sendx(3) when addrcnt is less than SCTP_SMALL_IOVEC_SIZE - re-add back inpcb_bind local address check bypass capability - Fix it so sctp_opt_info is independant of assoc_id postion. - Fix cookie life set to use MSEC_TO_TICKS() macro. - asconf changes o More comment changes/clarifications related to the old local address "not" list which is now an explicit restricted list. o Rename some functions for clarity: - sctp_add/del_local_addr_assoc to xxx_local_addr_restricted() - asconf related iterator functions to sctp_asconf_iterator_xxx() o Fix bug when the same address is deleted and added (and removed from the asconf queue) where the ifa is "freed" twice refcount wise, possibly freeing it completely. o Fix bug in output where the first ASCONF would not go out after the last address is changed (e.g. only goes out when retransmitted). o Fix bug where multiple ASCONFs can be bundled in the same packet with the and with the same serial numbers. o Fix asconf stcb iterator to not send ASCONF until after all work queue entries have been processed. o Change behavior so that when the last address is deleted (auto asconf on a bound all endpoint) no action is taken until an address is added; at that time, an ASCONF add+delete is sent (if the assoc is still up). o Fix local address counting so that address scoping is taken into account. o #ifdef SCTP_TIMER_BASED_ASCONF the old timer triggered sending of ASCONF (after an RTO). The default now is to send ASCONF immediately (except for the case of changing/deleting the last usable address). Approved by: re(ken smith)@freebsd.org
2007-07-24 20:06:02 +00:00
error = sctp_inpcb_bind(so, NULL, NULL, p);
if (error) {
goto out_now;
}
}
/* FIX ME: do we want to pass in a vrf on the connect call? */
vrf_id = inp->def_vrf_id;
/* We are GOOD to go */
stcb = sctp_aloc_assoc(inp, sa, &error, 0, vrf_id,
(struct thread *)p
);
if (stcb == NULL) {
/* Gak! no memory */
goto out_now;
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT);
/* move to second address */
switch (sa->sa_family) {
#ifdef INET
case AF_INET:
sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in));
break;
#endif
#ifdef INET6
case AF_INET6:
sa = (struct sockaddr *)((caddr_t)sa + sizeof(struct sockaddr_in6));
break;
#endif
default:
break;
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
error = 0;
added = sctp_connectx_helper_add(stcb, sa, (totaddr - 1), &error);
/* Fill in the return id */
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (error) {
(void)sctp_free_assoc(inp, stcb, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_6);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
goto out_now;
}
a_id = (sctp_assoc_t *) optval;
*a_id = sctp_get_associd(stcb);
/* initialize authentication parameters for the assoc */
sctp_initialize_auth_params(inp, stcb);
if (delay) {
/* doing delayed connection */
stcb->asoc.delayed_connection = 1;
sctp_timer_start(SCTP_TIMER_TYPE_INIT, inp, stcb, stcb->asoc.primary_destination);
} else {
(void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered);
sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED);
}
SCTP_TCB_UNLOCK(stcb);
if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED;
/* Set the connected flag so we can queue data */
soisconnecting(so);
}
out_now:
if (creat_lock_on) {
SCTP_ASOC_CREATE_UNLOCK(inp);
}
SCTP_INP_DECR_REF(inp);
return error;
}
#define SCTP_FIND_STCB(inp, stcb, assoc_id) { \
if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||\
(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) { \
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_RLOCK(inp); \
stcb = LIST_FIRST(&inp->sctp_asoc_list); \
if (stcb) { \
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_LOCK(stcb); \
} \
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_RUNLOCK(inp); \
} else if (assoc_id != 0) { \
stcb = sctp_findassociation_ep_asocid(inp, assoc_id, 1); \
if (stcb == NULL) { \
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT); \
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = ENOENT; \
break; \
} \
} else { \
stcb = NULL; \
} \
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
#define SCTP_CHECK_AND_CAST(destp, srcp, type, size) {\
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (size < sizeof(type)) { \
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL); \
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = EINVAL; \
break; \
} else { \
destp = (type *)srcp; \
} \
}
static int
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
sctp_getopt(struct socket *so, int optname, void *optval, size_t *optsize,
void *p)
{
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
struct sctp_inpcb *inp = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
int error, val = 0;
struct sctp_tcb *stcb = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (optval == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
return (EINVAL);
}
inp = (struct sctp_inpcb *)so->so_pcb;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (inp == 0) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
error = 0;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
switch (optname) {
case SCTP_NODELAY:
case SCTP_AUTOCLOSE:
case SCTP_EXPLICIT_EOR:
case SCTP_AUTO_ASCONF:
case SCTP_DISABLE_FRAGMENTS:
case SCTP_I_WANT_MAPPED_V4_ADDR:
case SCTP_USE_EXT_RCVINFO:
SCTP_INP_RLOCK(inp);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
switch (optname) {
case SCTP_DISABLE_FRAGMENTS:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NO_FRAGMENT);
break;
case SCTP_I_WANT_MAPPED_V4_ADDR:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4);
break;
case SCTP_AUTO_ASCONF:
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
/* only valid for bound all sockets */
val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
} else {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
goto flags_out;
}
break;
case SCTP_EXPLICIT_EOR:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXPLICIT_EOR);
break;
case SCTP_NODELAY:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NODELAY);
break;
case SCTP_USE_EXT_RCVINFO:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_EXT_RCVINFO);
break;
case SCTP_AUTOCLOSE:
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTOCLOSE))
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = TICKS_TO_SEC(inp->sctp_ep.auto_close_time);
else
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = 0;
break;
default:
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT);
error = ENOPROTOOPT;
} /* end switch (sopt->sopt_name) */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (optname != SCTP_AUTOCLOSE) {
/* make it an "on/off" value */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
val = (val != 0);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (*optsize < sizeof(val)) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
flags_out:
SCTP_INP_RUNLOCK(inp);
if (error == 0) {
/* return the option value */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*(int *)optval = val;
*optsize = sizeof(val);
}
break;
case SCTP_GET_PACKET_LOG:
{
#ifdef SCTP_PACKET_LOGGING
uint8_t *target;
int ret;
SCTP_CHECK_AND_CAST(target, optval, uint8_t, *optsize);
ret = sctp_copy_out_packet_log(target, (int)*optsize);
*optsize = ret;
#else
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
error = EOPNOTSUPP;
#endif
break;
}
case SCTP_REUSE_PORT:
{
uint32_t *value;
if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) {
/* Can't do this for a 1-m socket */
error = EINVAL;
break;
}
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
*value = sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE);
*optsize = sizeof(uint32_t);
}
break;
case SCTP_PARTIAL_DELIVERY_POINT:
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
uint32_t *value;
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
*value = inp->partial_delivery_point;
*optsize = sizeof(uint32_t);
}
break;
case SCTP_FRAGMENT_INTERLEAVE:
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
uint32_t *value;
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE)) {
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS)) {
*value = SCTP_FRAG_LEVEL_2;
} else {
*value = SCTP_FRAG_LEVEL_1;
}
} else {
*value = SCTP_FRAG_LEVEL_0;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(uint32_t);
}
break;
case SCTP_CMT_ON_OFF:
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
struct sctp_assoc_value *av;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
av->assoc_value = stcb->asoc.sctp_cmt_on_off;
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_RLOCK(inp);
av->assoc_value = inp->sctp_cmt_on_off;
SCTP_INP_RUNLOCK(inp);
}
*optsize = sizeof(*av);
}
break;
/* JRS - Get socket option for pluggable congestion control */
case SCTP_PLUGGABLE_CC:
{
struct sctp_assoc_value *av;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
av->assoc_value = stcb->asoc.congestion_control_module;
SCTP_TCB_UNLOCK(stcb);
} else {
av->assoc_value = inp->sctp_ep.sctp_default_cc_module;
}
*optsize = sizeof(*av);
}
break;
case SCTP_CC_OPTION:
{
struct sctp_cc_option *cc_opt;
SCTP_CHECK_AND_CAST(cc_opt, optval, struct sctp_cc_option, *optsize);
SCTP_FIND_STCB(inp, stcb, cc_opt->aid_value.assoc_id);
if (stcb == NULL) {
error = EINVAL;
} else {
if (stcb->asoc.cc_functions.sctp_cwnd_socket_option == NULL) {
error = ENOTSUP;
} else {
error = (*stcb->asoc.cc_functions.sctp_cwnd_socket_option) (stcb, 0,
cc_opt);
*optsize = sizeof(*cc_opt);
}
SCTP_TCB_UNLOCK(stcb);
}
}
break;
/* RS - Get socket option for pluggable stream scheduling */
case SCTP_PLUGGABLE_SS:
{
struct sctp_assoc_value *av;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
av->assoc_value = stcb->asoc.stream_scheduling_module;
SCTP_TCB_UNLOCK(stcb);
} else {
av->assoc_value = inp->sctp_ep.sctp_default_ss_module;
}
*optsize = sizeof(*av);
}
break;
case SCTP_SS_VALUE:
{
struct sctp_stream_value *av;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_stream_value, *optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
if (stcb->asoc.ss_functions.sctp_ss_get_value(stcb, &stcb->asoc, &stcb->asoc.strmout[av->stream_id],
&av->stream_value) < 0) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
} else {
*optsize = sizeof(*av);
}
SCTP_TCB_UNLOCK(stcb);
} else {
/*
* Can't get stream value without
* association
*/
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
}
break;
case SCTP_GET_ADDR_LEN:
{
struct sctp_assoc_value *av;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize);
error = EINVAL;
#ifdef INET
if (av->assoc_value == AF_INET) {
av->assoc_value = sizeof(struct sockaddr_in);
error = 0;
}
#endif
#ifdef INET6
if (av->assoc_value == AF_INET6) {
av->assoc_value = sizeof(struct sockaddr_in6);
error = 0;
}
#endif
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
if (error) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*av);
}
break;
case SCTP_GET_ASSOC_NUMBER:
{
uint32_t *value, cnt;
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
cnt = 0;
SCTP_INP_RLOCK(inp);
LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
cnt++;
}
SCTP_INP_RUNLOCK(inp);
*value = cnt;
*optsize = sizeof(uint32_t);
}
break;
case SCTP_GET_ASSOC_ID_LIST:
{
struct sctp_assoc_ids *ids;
unsigned int at, limit;
SCTP_CHECK_AND_CAST(ids, optval, struct sctp_assoc_ids, *optsize);
at = 0;
limit = (*optsize - sizeof(uint32_t)) / sizeof(sctp_assoc_t);
SCTP_INP_RLOCK(inp);
LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
if (at < limit) {
ids->gaids_assoc_id[at++] = sctp_get_associd(stcb);
} else {
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
break;
}
}
SCTP_INP_RUNLOCK(inp);
ids->gaids_number_of_ids = at;
*optsize = ((at * sizeof(sctp_assoc_t)) + sizeof(uint32_t));
}
break;
case SCTP_CONTEXT:
{
struct sctp_assoc_value *av;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
av->assoc_value = stcb->asoc.context;
SCTP_TCB_UNLOCK(stcb);
} else {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_RLOCK(inp);
av->assoc_value = inp->sctp_context;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_RUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*av);
}
break;
case SCTP_VRF_ID:
{
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
uint32_t *default_vrfid;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, *optsize);
*default_vrfid = inp->def_vrf_id;
break;
}
case SCTP_GET_ASOC_VRF:
{
struct sctp_assoc_value *id;
SCTP_CHECK_AND_CAST(id, optval, struct sctp_assoc_value, *optsize);
SCTP_FIND_STCB(inp, stcb, id->assoc_id);
if (stcb == NULL) {
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
break;
}
id->assoc_value = stcb->asoc.vrf_id;
break;
}
case SCTP_GET_VRF_IDS:
{
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
error = EOPNOTSUPP;
break;
}
case SCTP_GET_NONCE_VALUES:
{
struct sctp_get_nonce_values *gnv;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(gnv, optval, struct sctp_get_nonce_values, *optsize);
SCTP_FIND_STCB(inp, stcb, gnv->gn_assoc_id);
if (stcb) {
gnv->gn_peers_tag = stcb->asoc.peer_vtag;
gnv->gn_local_tag = stcb->asoc.my_vtag;
SCTP_TCB_UNLOCK(stcb);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = ENOTCONN;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*gnv);
}
break;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
case SCTP_DELAYED_SACK:
{
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
struct sctp_sack_info *sack;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, *optsize);
SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
sack->sack_delay = stcb->asoc.delayed_ack;
sack->sack_freq = stcb->asoc.sack_freq;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_RLOCK(inp);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
sack->sack_delay = TICKS_TO_MSEC(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV]);
sack->sack_freq = inp->sctp_ep.sctp_sack_freq;
SCTP_INP_RUNLOCK(inp);
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
*optsize = sizeof(*sack);
}
break;
case SCTP_GET_SNDBUF_USE:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
{
struct sctp_sockstat *ss;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(ss, optval, struct sctp_sockstat, *optsize);
SCTP_FIND_STCB(inp, stcb, ss->ss_assoc_id);
if (stcb) {
ss->ss_total_sndbuf = stcb->asoc.total_output_queue_size;
ss->ss_total_recv_buf = (stcb->asoc.size_on_reasm_queue +
stcb->asoc.size_on_all_streams);
SCTP_TCB_UNLOCK(stcb);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = ENOTCONN;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sctp_sockstat);
}
break;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
case SCTP_MAX_BURST:
{
struct sctp_assoc_value *av;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
av->assoc_value = stcb->asoc.max_burst;
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_RLOCK(inp);
av->assoc_value = inp->sctp_ep.max_burst;
SCTP_INP_RUNLOCK(inp);
}
*optsize = sizeof(struct sctp_assoc_value);
}
break;
case SCTP_MAXSEG:
{
struct sctp_assoc_value *av;
int ovh;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, *optsize);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
av->assoc_value = sctp_get_frag_point(stcb, &stcb->asoc);
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_RLOCK(inp);
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
ovh = SCTP_MED_OVERHEAD;
} else {
ovh = SCTP_MED_V4_OVERHEAD;
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (inp->sctp_frag_point >= SCTP_DEFAULT_MAXSEGMENT)
av->assoc_value = 0;
else
av->assoc_value = inp->sctp_frag_point - ovh;
SCTP_INP_RUNLOCK(inp);
}
*optsize = sizeof(struct sctp_assoc_value);
}
break;
case SCTP_GET_STAT_LOG:
error = sctp_fill_stat_log(optval, optsize);
break;
case SCTP_EVENTS:
{
struct sctp_event_subscribe *events;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, *optsize);
memset(events, 0, sizeof(*events));
SCTP_INP_RLOCK(inp);
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT))
events->sctp_data_io_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT))
events->sctp_association_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT))
events->sctp_address_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT))
events->sctp_send_failure_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR))
events->sctp_peer_error_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT))
events->sctp_shutdown_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT))
events->sctp_partial_delivery_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT))
events->sctp_adaptation_layer_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT))
events->sctp_authentication_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_DRYEVNT))
events->sctp_sender_dry_event = 1;
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT))
events->sctp_stream_reset_event = 1;
SCTP_INP_RUNLOCK(inp);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sctp_event_subscribe);
}
break;
case SCTP_ADAPTATION_LAYER:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
{
uint32_t *value;
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
SCTP_INP_RLOCK(inp);
*value = inp->sctp_ep.adaptation_layer_indicator;
SCTP_INP_RUNLOCK(inp);
*optsize = sizeof(uint32_t);
}
break;
case SCTP_SET_INITIAL_DBG_SEQ:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
{
uint32_t *value;
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
SCTP_INP_RLOCK(inp);
*value = inp->sctp_ep.initial_sequence_debug;
SCTP_INP_RUNLOCK(inp);
*optsize = sizeof(uint32_t);
}
break;
case SCTP_GET_LOCAL_ADDR_SIZE:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
{
uint32_t *value;
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
SCTP_INP_RLOCK(inp);
*value = sctp_count_max_addresses(inp);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_RUNLOCK(inp);
*optsize = sizeof(uint32_t);
}
break;
case SCTP_GET_REMOTE_ADDR_SIZE:
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
uint32_t *value;
size_t size;
struct sctp_nets *net;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(value, optval, uint32_t, *optsize);
/* FIXME MT: change to sctp_assoc_value? */
SCTP_FIND_STCB(inp, stcb, (sctp_assoc_t) * value);
if (stcb) {
size = 0;
/* Count the sizes */
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
size += sizeof(struct sockaddr_in6);
} else {
switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) {
#ifdef INET
case AF_INET:
size += sizeof(struct sockaddr_in);
break;
#endif
#ifdef INET6
case AF_INET6:
size += sizeof(struct sockaddr_in6);
break;
#endif
default:
break;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
}
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_UNLOCK(stcb);
*value = (uint32_t) size;
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = ENOTCONN;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(uint32_t);
}
break;
case SCTP_GET_PEER_ADDRESSES:
/*
* Get the address information, an array is passed in to
* fill up we pack it.
*/
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
size_t cpsz, left;
struct sockaddr_storage *sas;
struct sctp_nets *net;
struct sctp_getaddresses *saddr;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize);
SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
left = (*optsize) - sizeof(struct sctp_getaddresses);
*optsize = sizeof(struct sctp_getaddresses);
sas = (struct sockaddr_storage *)&saddr->addr[0];
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
cpsz = sizeof(struct sockaddr_in6);
} else {
switch (((struct sockaddr *)&net->ro._l_addr)->sa_family) {
#ifdef INET
case AF_INET:
cpsz = sizeof(struct sockaddr_in);
break;
#endif
#ifdef INET6
case AF_INET6:
cpsz = sizeof(struct sockaddr_in6);
break;
#endif
default:
cpsz = 0;
break;
}
}
if (cpsz == 0) {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
break;
}
if (left < cpsz) {
/* not enough room. */
break;
}
#if defined(INET) && defined(INET6)
if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_NEEDS_MAPPED_V4)) &&
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
(((struct sockaddr *)&net->ro._l_addr)->sa_family == AF_INET)) {
/* Must map the address */
in6_sin_2_v4mapsin6((struct sockaddr_in *)&net->ro._l_addr,
(struct sockaddr_in6 *)sas);
} else {
#endif
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
memcpy(sas, &net->ro._l_addr, cpsz);
#if defined(INET) && defined(INET6)
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
}
#endif
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
((struct sockaddr_in *)sas)->sin_port = stcb->rport;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
sas = (struct sockaddr_storage *)((caddr_t)sas + cpsz);
left -= cpsz;
*optsize += cpsz;
}
SCTP_TCB_UNLOCK(stcb);
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = ENOENT;
}
}
break;
case SCTP_GET_LOCAL_ADDRESSES:
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
size_t limit, actual;
struct sockaddr_storage *sas;
struct sctp_getaddresses *saddr;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(saddr, optval, struct sctp_getaddresses, *optsize);
SCTP_FIND_STCB(inp, stcb, saddr->sget_assoc_id);
sas = (struct sockaddr_storage *)&saddr->addr[0];
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
limit = *optsize - sizeof(sctp_assoc_t);
actual = sctp_fill_up_addresses(inp, stcb, limit, sas);
if (stcb) {
SCTP_TCB_UNLOCK(stcb);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sockaddr_storage) + actual;
}
break;
case SCTP_PEER_ADDR_PARAMS:
{
struct sctp_paddrparams *paddrp;
struct sctp_nets *net;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, *optsize);
SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id);
net = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
net = sctp_findnet(stcb, (struct sockaddr *)&paddrp->spp_address);
} else {
/*
* We increment here since
* sctp_findassociation_ep_addr() wil do a
* decrement if it finds the stcb as long as
* the locked tcb (last argument) is NOT a
* TCB.. aka NULL.
*/
SCTP_INP_INCR_REF(inp);
stcb = sctp_findassociation_ep_addr(&inp, (struct sockaddr *)&paddrp->spp_address, &net, NULL, NULL);
if (stcb == NULL) {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_DECR_REF(inp);
}
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (stcb && (net == NULL)) {
struct sockaddr *sa;
sa = (struct sockaddr *)&paddrp->spp_address;
#ifdef INET
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (sa->sa_family == AF_INET) {
struct sockaddr_in *sin;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
sin = (struct sockaddr_in *)sa;
if (sin->sin_addr.s_addr) {
error = EINVAL;
SCTP_TCB_UNLOCK(stcb);
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
break;
}
} else
#endif
#ifdef INET6
if (sa->sa_family == AF_INET6) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
struct sockaddr_in6 *sin6;
sin6 = (struct sockaddr_in6 *)sa;
if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
error = EINVAL;
SCTP_TCB_UNLOCK(stcb);
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
break;
}
} else
#endif
{
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
error = EAFNOSUPPORT;
SCTP_TCB_UNLOCK(stcb);
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
break;
}
}
if (stcb) {
/* Applys to the specific association */
paddrp->spp_flags = 0;
if (net) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
int ovh;
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
ovh = SCTP_MED_OVERHEAD;
} else {
ovh = SCTP_MED_V4_OVERHEAD;
}
paddrp->spp_pathmaxrxt = net->failure_threshold;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
paddrp->spp_pathmtu = net->mtu - ovh;
/* get flags for HB */
if (net->dest_state & SCTP_ADDR_NOHB)
paddrp->spp_flags |= SPP_HB_DISABLE;
else
paddrp->spp_flags |= SPP_HB_ENABLE;
/* get flags for PMTU */
if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) {
paddrp->spp_flags |= SPP_PMTUD_ENABLE;
} else {
paddrp->spp_flags |= SPP_PMTUD_DISABLE;
}
#ifdef INET
if (net->ro._l_addr.sin.sin_family == AF_INET) {
paddrp->spp_ipv4_tos = net->tos_flowlabel & 0x000000fc;
paddrp->spp_flags |= SPP_IPV4_TOS;
}
#endif
#ifdef INET6
if (net->ro._l_addr.sin6.sin6_family == AF_INET6) {
paddrp->spp_ipv6_flowlabel = net->tos_flowlabel;
paddrp->spp_flags |= SPP_IPV6_FLOWLABEL;
}
#endif
} else {
/*
* No destination so return default
* value
*/
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
int cnt = 0;
paddrp->spp_pathmaxrxt = stcb->asoc.def_net_failure;
paddrp->spp_pathmtu = sctp_get_frag_point(stcb, &stcb->asoc);
#ifdef INET
paddrp->spp_ipv4_tos = stcb->asoc.default_tos & 0x000000fc;
paddrp->spp_flags |= SPP_IPV4_TOS;
#endif
#ifdef INET6
paddrp->spp_ipv6_flowlabel = stcb->asoc.default_flowlabel;
paddrp->spp_flags |= SPP_IPV6_FLOWLABEL;
#endif
/* default settings should be these */
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (stcb->asoc.hb_is_disabled == 0) {
paddrp->spp_flags |= SPP_HB_ENABLE;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
} else {
paddrp->spp_flags |= SPP_HB_DISABLE;
}
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) {
cnt++;
}
}
if (cnt) {
paddrp->spp_flags |= SPP_PMTUD_ENABLE;
}
}
paddrp->spp_hbinterval = stcb->asoc.heart_beat_delay;
paddrp->spp_assoc_id = sctp_get_associd(stcb);
SCTP_TCB_UNLOCK(stcb);
} else {
/* Use endpoint defaults */
SCTP_INP_RLOCK(inp);
paddrp->spp_pathmaxrxt = inp->sctp_ep.def_net_failure;
paddrp->spp_hbinterval = TICKS_TO_MSEC(inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT]);
paddrp->spp_assoc_id = (sctp_assoc_t) 0;
/* get inp's default */
#ifdef INET
paddrp->spp_ipv4_tos = inp->ip_inp.inp.inp_ip_tos;
paddrp->spp_flags |= SPP_IPV4_TOS;
#endif
#ifdef INET6
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
paddrp->spp_ipv6_flowlabel = ((struct in6pcb *)inp)->in6p_flowinfo;
paddrp->spp_flags |= SPP_IPV6_FLOWLABEL;
}
#endif
/* can't return this */
paddrp->spp_pathmtu = 0;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
/* default behavior, no stcb */
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
paddrp->spp_flags = SPP_PMTUD_ENABLE;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
paddrp->spp_flags |= SPP_HB_ENABLE;
} else {
paddrp->spp_flags |= SPP_HB_DISABLE;
}
SCTP_INP_RUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sctp_paddrparams);
}
break;
case SCTP_GET_PEER_ADDR_INFO:
{
struct sctp_paddrinfo *paddri;
struct sctp_nets *net;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(paddri, optval, struct sctp_paddrinfo, *optsize);
SCTP_FIND_STCB(inp, stcb, paddri->spinfo_assoc_id);
net = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
net = sctp_findnet(stcb, (struct sockaddr *)&paddri->spinfo_address);
} else {
/*
* We increment here since
* sctp_findassociation_ep_addr() wil do a
* decrement if it finds the stcb as long as
* the locked tcb (last argument) is NOT a
* TCB.. aka NULL.
*/
SCTP_INP_INCR_REF(inp);
stcb = sctp_findassociation_ep_addr(&inp, (struct sockaddr *)&paddri->spinfo_address, &net, NULL, NULL);
if (stcb == NULL) {
SCTP_INP_DECR_REF(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if ((stcb) && (net)) {
if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
2011-01-20 13:53:34 +00:00
/* It's unconfirmed */
paddri->spinfo_state = SCTP_UNCONFIRMED;
} else if (net->dest_state & SCTP_ADDR_REACHABLE) {
2011-01-20 13:53:34 +00:00
/* It's active */
paddri->spinfo_state = SCTP_ACTIVE;
} else {
2011-01-20 13:53:34 +00:00
/* It's inactive */
paddri->spinfo_state = SCTP_INACTIVE;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
paddri->spinfo_cwnd = net->cwnd;
paddri->spinfo_srtt = net->lastsa >> SCTP_RTT_SHIFT;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
paddri->spinfo_rto = net->RTO;
paddri->spinfo_assoc_id = sctp_get_associd(stcb);
paddri->spinfo_mtu = net->mtu;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_UNLOCK(stcb);
} else {
if (stcb) {
SCTP_TCB_UNLOCK(stcb);
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
error = ENOENT;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sctp_paddrinfo);
}
break;
case SCTP_PCB_STATUS:
{
struct sctp_pcbinfo *spcb;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(spcb, optval, struct sctp_pcbinfo, *optsize);
sctp_fill_pcbinfo(spcb);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sctp_pcbinfo);
}
break;
case SCTP_STATUS:
{
struct sctp_nets *net;
struct sctp_status *sstat;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sstat, optval, struct sctp_status, *optsize);
SCTP_FIND_STCB(inp, stcb, sstat->sstat_assoc_id);
if (stcb == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
error = EINVAL;
break;
}
/*
* I think passing the state is fine since
* sctp_constants.h will be available to the user
* land.
*/
sstat->sstat_state = stcb->asoc.state;
sstat->sstat_assoc_id = sctp_get_associd(stcb);
sstat->sstat_rwnd = stcb->asoc.peers_rwnd;
sstat->sstat_unackdata = stcb->asoc.sent_queue_cnt;
/*
* We can't include chunks that have been passed to
* the socket layer. Only things in queue.
*/
sstat->sstat_penddata = (stcb->asoc.cnt_on_reasm_queue +
stcb->asoc.cnt_on_all_streams);
sstat->sstat_instrms = stcb->asoc.streamincnt;
sstat->sstat_outstrms = stcb->asoc.streamoutcnt;
sstat->sstat_fragmentation_point = sctp_get_frag_point(stcb, &stcb->asoc);
memcpy(&sstat->sstat_primary.spinfo_address,
&stcb->asoc.primary_destination->ro._l_addr,
((struct sockaddr *)(&stcb->asoc.primary_destination->ro._l_addr))->sa_len);
net = stcb->asoc.primary_destination;
((struct sockaddr_in *)&sstat->sstat_primary.spinfo_address)->sin_port = stcb->rport;
/*
* Again the user can get info from sctp_constants.h
* for what the state of the network is.
*/
if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
/* It's unconfirmed */
sstat->sstat_primary.spinfo_state = SCTP_UNCONFIRMED;
} else if (net->dest_state & SCTP_ADDR_REACHABLE) {
2011-01-20 13:53:34 +00:00
/* It's active */
sstat->sstat_primary.spinfo_state = SCTP_ACTIVE;
} else {
2011-01-20 13:53:34 +00:00
/* It's inactive */
sstat->sstat_primary.spinfo_state = SCTP_INACTIVE;
}
sstat->sstat_primary.spinfo_cwnd = net->cwnd;
sstat->sstat_primary.spinfo_srtt = net->lastsa >> SCTP_RTT_SHIFT;
sstat->sstat_primary.spinfo_rto = net->RTO;
sstat->sstat_primary.spinfo_mtu = net->mtu;
sstat->sstat_primary.spinfo_assoc_id = sctp_get_associd(stcb);
SCTP_TCB_UNLOCK(stcb);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*sstat);
}
break;
case SCTP_RTOINFO:
{
struct sctp_rtoinfo *srto;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, *optsize);
SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id);
if (stcb) {
srto->srto_initial = stcb->asoc.initial_rto;
srto->srto_max = stcb->asoc.maxrto;
srto->srto_min = stcb->asoc.minrto;
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_RLOCK(inp);
srto->srto_initial = inp->sctp_ep.initial_rto;
srto->srto_max = inp->sctp_ep.sctp_maxrto;
srto->srto_min = inp->sctp_ep.sctp_minrto;
SCTP_INP_RUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*srto);
}
break;
case SCTP_TIMEOUTS:
{
struct sctp_timeouts *stimo;
SCTP_CHECK_AND_CAST(stimo, optval, struct sctp_timeouts, *optsize);
SCTP_FIND_STCB(inp, stcb, stimo->stimo_assoc_id);
if (stcb) {
stimo->stimo_init = stcb->asoc.timoinit;
stimo->stimo_data = stcb->asoc.timodata;
stimo->stimo_sack = stcb->asoc.timosack;
stimo->stimo_shutdown = stcb->asoc.timoshutdown;
stimo->stimo_heartbeat = stcb->asoc.timoheartbeat;
stimo->stimo_cookie = stcb->asoc.timocookie;
stimo->stimo_shutdownack = stcb->asoc.timoshutdownack;
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
error = EINVAL;
}
*optsize = sizeof(*stimo);
}
break;
case SCTP_ASSOCINFO:
{
struct sctp_assocparams *sasoc;
uint32_t oldval;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, *optsize);
SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id);
if (stcb) {
oldval = sasoc->sasoc_cookie_life;
sasoc->sasoc_cookie_life = TICKS_TO_MSEC(stcb->asoc.cookie_life);
sasoc->sasoc_asocmaxrxt = stcb->asoc.max_send_times;
sasoc->sasoc_number_peer_destinations = stcb->asoc.numnets;
sasoc->sasoc_peer_rwnd = stcb->asoc.peers_rwnd;
sasoc->sasoc_local_rwnd = stcb->asoc.my_rwnd;
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_RLOCK(inp);
sasoc->sasoc_cookie_life = TICKS_TO_MSEC(inp->sctp_ep.def_cookie_life);
sasoc->sasoc_asocmaxrxt = inp->sctp_ep.max_send_times;
sasoc->sasoc_number_peer_destinations = 0;
sasoc->sasoc_peer_rwnd = 0;
sasoc->sasoc_local_rwnd = sbspace(&inp->sctp_socket->so_rcv);
SCTP_INP_RUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*sasoc);
}
break;
case SCTP_DEFAULT_SEND_PARAM:
{
struct sctp_sndrcvinfo *s_info;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, *optsize);
SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id);
if (stcb) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
memcpy(s_info, &stcb->asoc.def_send, sizeof(stcb->asoc.def_send));
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_RLOCK(inp);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
memcpy(s_info, &inp->def_send, sizeof(inp->def_send));
SCTP_INP_RUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*s_info);
}
break;
case SCTP_INITMSG:
{
struct sctp_initmsg *sinit;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, *optsize);
SCTP_INP_RLOCK(inp);
sinit->sinit_num_ostreams = inp->sctp_ep.pre_open_stream_count;
sinit->sinit_max_instreams = inp->sctp_ep.max_open_streams_intome;
sinit->sinit_max_attempts = inp->sctp_ep.max_init_times;
sinit->sinit_max_init_timeo = inp->sctp_ep.initial_init_rto_max;
SCTP_INP_RUNLOCK(inp);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*sinit);
}
break;
case SCTP_PRIMARY_ADDR:
/* we allow a "get" operation on this */
{
struct sctp_setprim *ssp;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(ssp, optval, struct sctp_setprim, *optsize);
SCTP_FIND_STCB(inp, stcb, ssp->ssp_assoc_id);
if (stcb) {
/* simply copy out the sockaddr_storage... */
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
int len;
len = *optsize;
if (len > stcb->asoc.primary_destination->ro._l_addr.sa.sa_len)
len = stcb->asoc.primary_destination->ro._l_addr.sa.sa_len;
memcpy(&ssp->ssp_addr,
&stcb->asoc.primary_destination->ro._l_addr,
len);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_UNLOCK(stcb);
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = EINVAL;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*ssp);
}
break;
case SCTP_HMAC_IDENT:
{
struct sctp_hmacalgo *shmac;
sctp_hmaclist_t *hmaclist;
uint32_t size;
int i;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, *optsize);
SCTP_INP_RLOCK(inp);
hmaclist = inp->sctp_ep.local_hmacs;
if (hmaclist == NULL) {
/* no HMACs to return */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*shmac);
SCTP_INP_RUNLOCK(inp);
break;
}
/* is there room for all of the hmac ids? */
size = sizeof(*shmac) + (hmaclist->num_algo *
sizeof(shmac->shmac_idents[0]));
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if ((size_t)(*optsize) < size) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
error = EINVAL;
SCTP_INP_RUNLOCK(inp);
break;
}
/* copy in the list */
shmac->shmac_number_of_idents = hmaclist->num_algo;
for (i = 0; i < hmaclist->num_algo; i++) {
shmac->shmac_idents[i] = hmaclist->hmac[i];
}
SCTP_INP_RUNLOCK(inp);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = size;
break;
}
case SCTP_AUTH_ACTIVE_KEY:
{
struct sctp_authkeyid *scact;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid, *optsize);
SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id);
if (stcb) {
/* get the active key on the assoc */
scact->scact_keynumber = stcb->asoc.authinfo.active_keyid;
SCTP_TCB_UNLOCK(stcb);
} else {
/* get the endpoint active key */
SCTP_INP_RLOCK(inp);
scact->scact_keynumber = inp->sctp_ep.default_keyid;
SCTP_INP_RUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(*scact);
break;
}
case SCTP_LOCAL_AUTH_CHUNKS:
{
struct sctp_authchunks *sac;
sctp_auth_chklist_t *chklist = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
size_t size = 0;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize);
SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id);
if (stcb) {
/* get off the assoc */
chklist = stcb->asoc.local_auth_chunks;
/* is there enough space? */
size = sctp_auth_get_chklist_size(chklist);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (*optsize < (sizeof(struct sctp_authchunks) + size)) {
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
/* copy in the chunks */
(void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks);
}
SCTP_TCB_UNLOCK(stcb);
} else {
/* get off the endpoint */
SCTP_INP_RLOCK(inp);
chklist = inp->sctp_ep.local_auth_chunks;
/* is there enough space? */
size = sctp_auth_get_chklist_size(chklist);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (*optsize < (sizeof(struct sctp_authchunks) + size)) {
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
/* copy in the chunks */
(void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks);
}
SCTP_INP_RUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sctp_authchunks) + size;
break;
}
case SCTP_PEER_AUTH_CHUNKS:
{
struct sctp_authchunks *sac;
sctp_auth_chklist_t *chklist = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
size_t size = 0;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sac, optval, struct sctp_authchunks, *optsize);
SCTP_FIND_STCB(inp, stcb, sac->gauth_assoc_id);
if (stcb) {
/* get off the assoc */
chklist = stcb->asoc.peer_auth_chunks;
/* is there enough space? */
size = sctp_auth_get_chklist_size(chklist);
if (*optsize < (sizeof(struct sctp_authchunks) + size)) {
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
/* copy in the chunks */
(void)sctp_serialize_auth_chunks(chklist, sac->gauth_chunks);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
}
SCTP_TCB_UNLOCK(stcb);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = ENOENT;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = sizeof(struct sctp_authchunks) + size;
break;
}
default:
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT);
error = ENOPROTOOPT;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
*optsize = 0;
break;
} /* end switch (sopt->sopt_name) */
return (error);
}
static int
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
sctp_setopt(struct socket *so, int optname, void *optval, size_t optsize,
void *p)
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
int error, set_opt;
uint32_t *mopt;
struct sctp_tcb *stcb = NULL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
struct sctp_inpcb *inp = NULL;
uint32_t vrf_id;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (optval == NULL) {
SCTP_PRINTF("optval is NULL\n");
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (EINVAL);
}
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0) {
SCTP_PRINTF("inp is NULL?\n");
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return EINVAL;
}
vrf_id = inp->def_vrf_id;
error = 0;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
switch (optname) {
case SCTP_NODELAY:
case SCTP_AUTOCLOSE:
case SCTP_AUTO_ASCONF:
case SCTP_EXPLICIT_EOR:
case SCTP_DISABLE_FRAGMENTS:
case SCTP_USE_EXT_RCVINFO:
case SCTP_I_WANT_MAPPED_V4_ADDR:
/* copy in the option value */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize);
set_opt = 0;
if (error)
break;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
switch (optname) {
case SCTP_DISABLE_FRAGMENTS:
set_opt = SCTP_PCB_FLAGS_NO_FRAGMENT;
break;
case SCTP_AUTO_ASCONF:
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
/*
* NOTE: we don't really support this flag
*/
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
/* only valid for bound all sockets */
set_opt = SCTP_PCB_FLAGS_AUTO_ASCONF;
} else {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (EINVAL);
}
break;
case SCTP_EXPLICIT_EOR:
set_opt = SCTP_PCB_FLAGS_EXPLICIT_EOR;
break;
case SCTP_USE_EXT_RCVINFO:
set_opt = SCTP_PCB_FLAGS_EXT_RCVINFO;
break;
case SCTP_I_WANT_MAPPED_V4_ADDR:
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
set_opt = SCTP_PCB_FLAGS_NEEDS_MAPPED_V4;
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (EINVAL);
}
break;
case SCTP_NODELAY:
set_opt = SCTP_PCB_FLAGS_NODELAY;
break;
case SCTP_AUTOCLOSE:
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
return (EINVAL);
}
set_opt = SCTP_PCB_FLAGS_AUTOCLOSE;
/*
* The value is in ticks. Note this does not effect
* old associations, only new ones.
*/
inp->sctp_ep.auto_close_time = SEC_TO_TICKS(*mopt);
break;
}
SCTP_INP_WLOCK(inp);
if (*mopt != 0) {
sctp_feature_on(inp, set_opt);
} else {
sctp_feature_off(inp, set_opt);
}
SCTP_INP_WUNLOCK(inp);
break;
case SCTP_REUSE_PORT:
{
SCTP_CHECK_AND_CAST(mopt, optval, uint32_t, optsize);
if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
/* Can't set it after we are bound */
error = EINVAL;
break;
}
if ((inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE)) {
/* Can't do this for a 1-m socket */
error = EINVAL;
break;
}
if (optval)
sctp_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE);
else
sctp_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE);
}
break;
case SCTP_PARTIAL_DELIVERY_POINT:
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
uint32_t *value;
SCTP_CHECK_AND_CAST(value, optval, uint32_t, optsize);
if (*value > SCTP_SB_LIMIT_RCV(so)) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
inp->partial_delivery_point = *value;
}
break;
case SCTP_FRAGMENT_INTERLEAVE:
/* not yet until we re-write sctp_recvmsg() */
{
uint32_t *level;
SCTP_CHECK_AND_CAST(level, optval, uint32_t, optsize);
if (*level == SCTP_FRAG_LEVEL_2) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
} else if (*level == SCTP_FRAG_LEVEL_1) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
} else if (*level == SCTP_FRAG_LEVEL_0) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
}
break;
case SCTP_CMT_ON_OFF:
if (SCTP_BASE_SYSCTL(sctp_cmt_on_off)) {
struct sctp_assoc_value *av;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
if (av->assoc_value > SCTP_CMT_MAX) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
} else {
stcb->asoc.sctp_cmt_on_off = av->assoc_value;
}
SCTP_TCB_UNLOCK(stcb);
} else {
if (av->assoc_value > SCTP_CMT_MAX) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
} else {
SCTP_INP_WLOCK(inp);
inp->sctp_cmt_on_off = av->assoc_value;
SCTP_INP_WUNLOCK(inp);
}
}
} else {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT);
error = ENOPROTOOPT;
}
break;
/* JRS - Set socket option for pluggable congestion control */
case SCTP_PLUGGABLE_CC:
{
struct sctp_assoc_value *av;
struct sctp_nets *net;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
switch (av->assoc_value) {
case SCTP_CC_RFC2581:
case SCTP_CC_HSTCP:
case SCTP_CC_HTCP:
case SCTP_CC_RTCC:
stcb->asoc.cc_functions = sctp_cc_functions[av->assoc_value];
stcb->asoc.congestion_control_module = av->assoc_value;
if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL) {
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net);
}
}
break;
default:
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
SCTP_TCB_UNLOCK(stcb);
} else {
switch (av->assoc_value) {
case SCTP_CC_RFC2581:
case SCTP_CC_HSTCP:
case SCTP_CC_HTCP:
case SCTP_CC_RTCC:
SCTP_INP_WLOCK(inp);
inp->sctp_ep.sctp_default_cc_module = av->assoc_value;
SCTP_INP_WUNLOCK(inp);
break;
default:
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
}
}
break;
case SCTP_CC_OPTION:
{
struct sctp_cc_option *cc_opt;
SCTP_CHECK_AND_CAST(cc_opt, optval, struct sctp_cc_option, optsize);
SCTP_FIND_STCB(inp, stcb, cc_opt->aid_value.assoc_id);
if (stcb == NULL) {
error = EINVAL;
} else {
if (stcb->asoc.cc_functions.sctp_cwnd_socket_option == NULL) {
error = ENOTSUP;
} else {
error = (*stcb->asoc.cc_functions.sctp_cwnd_socket_option) (stcb, 1,
cc_opt);
}
SCTP_TCB_UNLOCK(stcb);
}
}
break;
/* RS - Set socket option for pluggable stream scheduling */
case SCTP_PLUGGABLE_SS:
{
struct sctp_assoc_value *av;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
switch (av->assoc_value) {
case SCTP_SS_DEFAULT:
case SCTP_SS_ROUND_ROBIN:
case SCTP_SS_ROUND_ROBIN_PACKET:
case SCTP_SS_PRIORITY:
case SCTP_SS_FAIR_BANDWITH:
case SCTP_SS_FIRST_COME:
stcb->asoc.ss_functions.sctp_ss_clear(stcb, &stcb->asoc, 1, 1);
stcb->asoc.ss_functions = sctp_ss_functions[av->assoc_value];
stcb->asoc.stream_scheduling_module = av->assoc_value;
stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 1);
break;
default:
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
SCTP_TCB_UNLOCK(stcb);
} else {
switch (av->assoc_value) {
case SCTP_SS_DEFAULT:
case SCTP_SS_ROUND_ROBIN:
case SCTP_SS_ROUND_ROBIN_PACKET:
case SCTP_SS_PRIORITY:
case SCTP_SS_FAIR_BANDWITH:
case SCTP_SS_FIRST_COME:
SCTP_INP_WLOCK(inp);
inp->sctp_ep.sctp_default_ss_module = av->assoc_value;
SCTP_INP_WUNLOCK(inp);
break;
default:
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
}
}
break;
case SCTP_SS_VALUE:
{
struct sctp_stream_value *av;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_stream_value, optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
if (stcb->asoc.ss_functions.sctp_ss_set_value(stcb, &stcb->asoc, &stcb->asoc.strmout[av->stream_id],
av->stream_value) < 0) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
SCTP_TCB_UNLOCK(stcb);
} else {
/*
* Can't set stream value without
* association
*/
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
}
break;
case SCTP_CLR_STAT_LOG:
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
error = EOPNOTSUPP;
break;
case SCTP_CONTEXT:
{
struct sctp_assoc_value *av;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
if (stcb) {
stcb->asoc.context = av->assoc_value;
SCTP_TCB_UNLOCK(stcb);
} else {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_WLOCK(inp);
inp->sctp_context = av->assoc_value;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_WUNLOCK(inp);
}
}
break;
case SCTP_VRF_ID:
{
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
uint32_t *default_vrfid;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
SCTP_CHECK_AND_CAST(default_vrfid, optval, uint32_t, optsize);
if (*default_vrfid > SCTP_MAX_VRF_ID) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
inp->def_vrf_id = *default_vrfid;
break;
}
case SCTP_DEL_VRF_ID:
{
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
error = EOPNOTSUPP;
break;
}
case SCTP_ADD_VRF_ID:
{
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
error = EOPNOTSUPP;
break;
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
case SCTP_DELAYED_SACK:
{
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
struct sctp_sack_info *sack;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
SCTP_CHECK_AND_CAST(sack, optval, struct sctp_sack_info, optsize);
SCTP_FIND_STCB(inp, stcb, sack->sack_assoc_id);
if (sack->sack_delay) {
if (sack->sack_delay > SCTP_MAX_SACK_DELAY)
sack->sack_delay = SCTP_MAX_SACK_DELAY;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (sack->sack_delay) {
if (MSEC_TO_TICKS(sack->sack_delay) < 1) {
sack->sack_delay = TICKS_TO_MSEC(1);
}
stcb->asoc.delayed_ack = sack->sack_delay;
}
if (sack->sack_freq) {
stcb->asoc.sack_freq = sack->sack_freq;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_WLOCK(inp);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (sack->sack_delay) {
if (MSEC_TO_TICKS(sack->sack_delay) < 1) {
sack->sack_delay = TICKS_TO_MSEC(1);
}
inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(sack->sack_delay);
}
if (sack->sack_freq) {
inp->sctp_ep.sctp_sack_freq = sack->sack_freq;
}
SCTP_INP_WUNLOCK(inp);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
break;
}
case SCTP_AUTH_CHUNK:
{
struct sctp_authchunk *sauth;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sauth, optval, struct sctp_authchunk, optsize);
SCTP_INP_WLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (sctp_auth_add_chunk(sauth->sauth_chunk, inp->sctp_ep.local_auth_chunks)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_WUNLOCK(inp);
break;
}
case SCTP_AUTH_KEY:
{
struct sctp_authkey *sca;
struct sctp_keyhead *shared_keys;
sctp_sharedkey_t *shared_key;
sctp_key_t *key = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
size_t size;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sca, optval, struct sctp_authkey, optsize);
SCTP_FIND_STCB(inp, stcb, sca->sca_assoc_id);
size = optsize - sizeof(*sca);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
/* set it on the assoc */
shared_keys = &stcb->asoc.shared_keys;
/* clear the cached keys for this key id */
sctp_clear_cachedkeys(stcb, sca->sca_keynumber);
/*
* create the new shared key and
* insert/replace it
*/
if (size > 0) {
key = sctp_set_key(sca->sca_key, (uint32_t) size);
if (key == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM);
error = ENOMEM;
SCTP_TCB_UNLOCK(stcb);
break;
}
}
shared_key = sctp_alloc_sharedkey();
if (shared_key == NULL) {
sctp_free_key(key);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM);
error = ENOMEM;
SCTP_TCB_UNLOCK(stcb);
break;
}
shared_key->key = key;
shared_key->keyid = sca->sca_keynumber;
error = sctp_insert_sharedkey(shared_keys, shared_key);
SCTP_TCB_UNLOCK(stcb);
} else {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
/* set it on the endpoint */
SCTP_INP_WLOCK(inp);
shared_keys = &inp->sctp_ep.shared_keys;
/*
* clear the cached keys on all assocs for
* this key id
*/
sctp_clear_cachedkeys_ep(inp, sca->sca_keynumber);
/*
* create the new shared key and
* insert/replace it
*/
if (size > 0) {
key = sctp_set_key(sca->sca_key, (uint32_t) size);
if (key == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM);
error = ENOMEM;
SCTP_INP_WUNLOCK(inp);
break;
}
}
shared_key = sctp_alloc_sharedkey();
if (shared_key == NULL) {
sctp_free_key(key);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM);
error = ENOMEM;
SCTP_INP_WUNLOCK(inp);
break;
}
shared_key->key = key;
shared_key->keyid = sca->sca_keynumber;
error = sctp_insert_sharedkey(shared_keys, shared_key);
SCTP_INP_WUNLOCK(inp);
}
break;
}
case SCTP_HMAC_IDENT:
{
struct sctp_hmacalgo *shmac;
sctp_hmaclist_t *hmaclist;
uint16_t hmacid;
uint32_t i;
size_t found;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(shmac, optval, struct sctp_hmacalgo, optsize);
if (optsize < sizeof(struct sctp_hmacalgo) + shmac->shmac_number_of_idents * sizeof(uint16_t)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
hmaclist = sctp_alloc_hmaclist(shmac->shmac_number_of_idents);
if (hmaclist == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOMEM);
error = ENOMEM;
break;
}
for (i = 0; i < shmac->shmac_number_of_idents; i++) {
hmacid = shmac->shmac_idents[i];
if (sctp_auth_add_hmacid(hmaclist, hmacid)) {
/* invalid HMACs were found */ ;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
sctp_free_hmaclist(hmaclist);
goto sctp_set_hmac_done;
}
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
found = 0;
for (i = 0; i < hmaclist->num_algo; i++) {
if (hmaclist->hmac[i] == SCTP_AUTH_HMAC_ID_SHA1) {
/* already in list */
found = 1;
}
}
if (!found) {
sctp_free_hmaclist(hmaclist);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
error = EINVAL;
break;
}
/* set it on the endpoint */
SCTP_INP_WLOCK(inp);
if (inp->sctp_ep.local_hmacs)
sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
inp->sctp_ep.local_hmacs = hmaclist;
SCTP_INP_WUNLOCK(inp);
sctp_set_hmac_done:
break;
}
case SCTP_AUTH_ACTIVE_KEY:
{
struct sctp_authkeyid *scact;
SCTP_CHECK_AND_CAST(scact, optval, struct sctp_authkeyid,
optsize);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_FIND_STCB(inp, stcb, scact->scact_assoc_id);
/* set the active key on the right place */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
/* set the active key on the assoc */
if (sctp_auth_setactivekey(stcb,
scact->scact_keynumber)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL,
SCTP_FROM_SCTP_USRREQ,
EINVAL);
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
SCTP_TCB_UNLOCK(stcb);
} else {
/* set the active key on the endpoint */
SCTP_INP_WLOCK(inp);
if (sctp_auth_setactivekey_ep(inp,
scact->scact_keynumber)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL,
SCTP_FROM_SCTP_USRREQ,
EINVAL);
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
SCTP_INP_WUNLOCK(inp);
}
break;
}
case SCTP_AUTH_DELETE_KEY:
{
struct sctp_authkeyid *scdel;
SCTP_CHECK_AND_CAST(scdel, optval, struct sctp_authkeyid,
optsize);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_FIND_STCB(inp, stcb, scdel->scact_assoc_id);
/* delete the key from the right place */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
if (sctp_delete_sharedkey(stcb,
scdel->scact_keynumber)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL,
SCTP_FROM_SCTP_USRREQ,
EINVAL);
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_WLOCK(inp);
if (sctp_delete_sharedkey_ep(inp,
scdel->scact_keynumber)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL,
SCTP_FROM_SCTP_USRREQ,
EINVAL);
error = EINVAL;
}
SCTP_INP_WUNLOCK(inp);
}
break;
}
case SCTP_AUTH_DEACTIVATE_KEY:
{
struct sctp_authkeyid *keyid;
SCTP_CHECK_AND_CAST(keyid, optval, struct sctp_authkeyid,
optsize);
SCTP_FIND_STCB(inp, stcb, keyid->scact_assoc_id);
/* deactivate the key from the right place */
if (stcb) {
if (sctp_deact_sharedkey(stcb,
keyid->scact_keynumber)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL,
SCTP_FROM_SCTP_USRREQ,
EINVAL);
error = EINVAL;
}
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_WLOCK(inp);
if (sctp_deact_sharedkey_ep(inp,
keyid->scact_keynumber)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL,
SCTP_FROM_SCTP_USRREQ,
EINVAL);
error = EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
SCTP_INP_WUNLOCK(inp);
}
break;
}
case SCTP_RESET_STREAMS:
{
struct sctp_stream_reset *strrst;
uint8_t send_in = 0, send_tsn = 0, send_out = 0,
addstream = 0;
uint16_t addstrmcnt = 0;
int i;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(strrst, optval, struct sctp_stream_reset, optsize);
SCTP_FIND_STCB(inp, stcb, strrst->strrst_assoc_id);
if (stcb == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
error = ENOENT;
break;
}
if (stcb->asoc.peer_supports_strreset == 0) {
/*
* Peer does not support it, we return
* protocol not supported since this is true
* for this feature and this peer, not the
* socket request in general.
*/
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EPROTONOSUPPORT);
error = EPROTONOSUPPORT;
SCTP_TCB_UNLOCK(stcb);
break;
}
if (stcb->asoc.stream_reset_outstanding) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY);
error = EALREADY;
SCTP_TCB_UNLOCK(stcb);
break;
}
if (strrst->strrst_flags == SCTP_RESET_LOCAL_RECV) {
send_in = 1;
} else if (strrst->strrst_flags == SCTP_RESET_LOCAL_SEND) {
send_out = 1;
} else if (strrst->strrst_flags == SCTP_RESET_BOTH) {
send_in = 1;
send_out = 1;
} else if (strrst->strrst_flags == SCTP_RESET_TSN) {
send_tsn = 1;
} else if (strrst->strrst_flags == SCTP_RESET_ADD_STREAMS) {
if (send_tsn ||
send_in ||
send_out) {
/* We can't do that and add streams */
error = EINVAL;
goto skip_stuff;
}
if (stcb->asoc.stream_reset_outstanding) {
error = EBUSY;
goto skip_stuff;
}
addstream = 1;
/* We allocate here */
addstrmcnt = strrst->strrst_num_streams;
if ((int)(addstrmcnt + stcb->asoc.streamoutcnt) > 0xffff) {
/* You can't have more than 64k */
error = EINVAL;
goto skip_stuff;
}
if ((stcb->asoc.strm_realoutsize - stcb->asoc.streamoutcnt) < addstrmcnt) {
/* Need to allocate more */
struct sctp_stream_out *oldstream;
struct sctp_stream_queue_pending *sp,
*nsp;
oldstream = stcb->asoc.strmout;
/* get some more */
SCTP_MALLOC(stcb->asoc.strmout, struct sctp_stream_out *,
((stcb->asoc.streamoutcnt + addstrmcnt) * sizeof(struct sctp_stream_out)),
SCTP_M_STRMO);
if (stcb->asoc.strmout == NULL) {
stcb->asoc.strmout = oldstream;
error = ENOMEM;
goto skip_stuff;
}
/*
* Ok now we proceed with copying
* the old out stuff and
* initializing the new stuff.
*/
SCTP_TCB_SEND_LOCK(stcb);
stcb->asoc.ss_functions.sctp_ss_clear(stcb, &stcb->asoc, 0, 1);
for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
TAILQ_INIT(&stcb->asoc.strmout[i].outqueue);
stcb->asoc.strmout[i].next_sequence_sent = oldstream[i].next_sequence_sent;
stcb->asoc.strmout[i].last_msg_incomplete = oldstream[i].last_msg_incomplete;
stcb->asoc.strmout[i].stream_no = i;
stcb->asoc.ss_functions.sctp_ss_init_stream(&stcb->asoc.strmout[i], &oldstream[i]);
/*
* now anything on those
* queues?
*/
TAILQ_FOREACH_SAFE(sp, &oldstream[i].outqueue, next, nsp) {
TAILQ_REMOVE(&oldstream[i].outqueue, sp, next);
TAILQ_INSERT_TAIL(&stcb->asoc.strmout[i].outqueue, sp, next);
}
/*
* Now move assoc pointers
* too
*/
if (stcb->asoc.last_out_stream == &oldstream[i]) {
stcb->asoc.last_out_stream = &stcb->asoc.strmout[i];
}
if (stcb->asoc.locked_on_sending == &oldstream[i]) {
stcb->asoc.locked_on_sending = &stcb->asoc.strmout[i];
}
}
/* now the new streams */
stcb->asoc.ss_functions.sctp_ss_init(stcb, &stcb->asoc, 1);
for (i = stcb->asoc.streamoutcnt; i < (stcb->asoc.streamoutcnt + addstrmcnt); i++) {
stcb->asoc.strmout[i].next_sequence_sent = 0x0;
TAILQ_INIT(&stcb->asoc.strmout[i].outqueue);
stcb->asoc.strmout[i].stream_no = i;
stcb->asoc.strmout[i].last_msg_incomplete = 0;
stcb->asoc.ss_functions.sctp_ss_init_stream(&stcb->asoc.strmout[i], NULL);
}
stcb->asoc.strm_realoutsize = stcb->asoc.streamoutcnt + addstrmcnt;
SCTP_FREE(oldstream, SCTP_M_STRMO);
}
SCTP_TCB_SEND_UNLOCK(stcb);
goto skip_stuff;
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
SCTP_TCB_UNLOCK(stcb);
break;
}
for (i = 0; i < strrst->strrst_num_streams; i++) {
if ((send_in) &&
(strrst->strrst_list[i] > stcb->asoc.streamincnt)) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
goto get_out;
}
if ((send_out) &&
(strrst->strrst_list[i] > stcb->asoc.streamoutcnt)) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
goto get_out;
}
}
skip_stuff:
if (error) {
get_out:
SCTP_TCB_UNLOCK(stcb);
break;
}
error = sctp_send_str_reset_req(stcb, strrst->strrst_num_streams,
strrst->strrst_list,
send_out, (stcb->asoc.str_reset_seq_in - 3),
send_in, send_tsn, addstream, addstrmcnt);
sctp_chunk_output(inp, stcb, SCTP_OUTPUT_FROM_STRRST_REQ, SCTP_SO_LOCKED);
SCTP_TCB_UNLOCK(stcb);
}
break;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
case SCTP_CONNECT_X:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = sctp_do_connect_x(so, inp, optval, optsize, p, 0);
break;
case SCTP_CONNECT_X_DELAYED:
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (optsize < (sizeof(int) + sizeof(struct sockaddr_in))) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = sctp_do_connect_x(so, inp, optval, optsize, p, 1);
break;
case SCTP_CONNECT_X_COMPLETE:
{
struct sockaddr *sa;
struct sctp_nets *net;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
/* FIXME MT: check correct? */
SCTP_CHECK_AND_CAST(sa, optval, struct sockaddr, optsize);
/* find tcb */
if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
SCTP_INP_RLOCK(inp);
stcb = LIST_FIRST(&inp->sctp_asoc_list);
if (stcb) {
SCTP_TCB_LOCK(stcb);
net = sctp_findnet(stcb, sa);
}
SCTP_INP_RUNLOCK(inp);
} else {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
/*
* We increment here since
* sctp_findassociation_ep_addr() wil do a
* decrement if it finds the stcb as long as
* the locked tcb (last argument) is NOT a
* TCB.. aka NULL.
*/
SCTP_INP_INCR_REF(inp);
stcb = sctp_findassociation_ep_addr(&inp, sa, &net, NULL, NULL);
if (stcb == NULL) {
SCTP_INP_DECR_REF(inp);
}
}
if (stcb == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
error = ENOENT;
break;
}
if (stcb->asoc.delayed_connection == 1) {
stcb->asoc.delayed_connection = 0;
(void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered);
sctp_timer_stop(SCTP_TIMER_TYPE_INIT, inp, stcb,
stcb->asoc.primary_destination,
SCTP_FROM_SCTP_USRREQ + SCTP_LOC_9);
sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED);
} else {
/*
* already expired or did not use delayed
* connectx
*/
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY);
error = EALREADY;
}
SCTP_TCB_UNLOCK(stcb);
}
break;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
case SCTP_MAX_BURST:
{
struct sctp_assoc_value *av;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
stcb->asoc.max_burst = av->assoc_value;
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_WLOCK(inp);
inp->sctp_ep.max_burst = av->assoc_value;
SCTP_INP_WUNLOCK(inp);
}
}
break;
case SCTP_MAXSEG:
{
struct sctp_assoc_value *av;
int ovh;
SCTP_CHECK_AND_CAST(av, optval, struct sctp_assoc_value, optsize);
SCTP_FIND_STCB(inp, stcb, av->assoc_id);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
ovh = SCTP_MED_OVERHEAD;
} else {
ovh = SCTP_MED_V4_OVERHEAD;
}
if (stcb) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (av->assoc_value) {
stcb->asoc.sctp_frag_point = (av->assoc_value + ovh);
} else {
stcb->asoc.sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
}
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_WLOCK(inp);
/*
* FIXME MT: I think this is not in tune
* with the API ID
*/
if (av->assoc_value) {
inp->sctp_frag_point = (av->assoc_value + ovh);
} else {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
}
SCTP_INP_WUNLOCK(inp);
}
}
break;
case SCTP_EVENTS:
{
struct sctp_event_subscribe *events;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(events, optval, struct sctp_event_subscribe, optsize);
SCTP_INP_WLOCK(inp);
if (events->sctp_data_io_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVDATAIOEVNT);
}
if (events->sctp_association_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVASSOCEVNT);
}
if (events->sctp_address_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPADDREVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPADDREVNT);
}
if (events->sctp_send_failure_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSENDFAILEVNT);
}
if (events->sctp_peer_error_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVPEERERR);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVPEERERR);
}
if (events->sctp_shutdown_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_RECVSHUTDOWNEVNT);
}
if (events->sctp_partial_delivery_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_PDAPIEVNT);
}
if (events->sctp_adaptation_layer_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_ADAPTATIONEVNT);
}
if (events->sctp_authentication_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTHEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTHEVNT);
}
if (events->sctp_sender_dry_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_DRYEVNT);
if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
stcb = LIST_FIRST(&inp->sctp_asoc_list);
if (stcb) {
SCTP_TCB_LOCK(stcb);
}
if (stcb &&
TAILQ_EMPTY(&stcb->asoc.send_queue) &&
TAILQ_EMPTY(&stcb->asoc.sent_queue) &&
(stcb->asoc.stream_queue_cnt == 0)) {
sctp_ulp_notify(SCTP_NOTIFY_SENDER_DRY, stcb, 0, NULL, SCTP_SO_LOCKED);
}
if (stcb) {
SCTP_TCB_UNLOCK(stcb);
}
}
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_DRYEVNT);
}
if (events->sctp_stream_reset_event) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT);
} else {
sctp_feature_off(inp, SCTP_PCB_FLAGS_STREAM_RESETEVNT);
}
SCTP_INP_WUNLOCK(inp);
}
break;
case SCTP_ADAPTATION_LAYER:
{
struct sctp_setadaptation *adap_bits;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(adap_bits, optval, struct sctp_setadaptation, optsize);
SCTP_INP_WLOCK(inp);
inp->sctp_ep.adaptation_layer_indicator = adap_bits->ssb_adaptation_ind;
SCTP_INP_WUNLOCK(inp);
}
break;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
#ifdef SCTP_DEBUG
case SCTP_SET_INITIAL_DBG_SEQ:
{
uint32_t *vvv;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(vvv, optval, uint32_t, optsize);
SCTP_INP_WLOCK(inp);
inp->sctp_ep.initial_sequence_debug = *vvv;
SCTP_INP_WUNLOCK(inp);
}
break;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
#endif
case SCTP_DEFAULT_SEND_PARAM:
{
struct sctp_sndrcvinfo *s_info;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(s_info, optval, struct sctp_sndrcvinfo, optsize);
SCTP_FIND_STCB(inp, stcb, s_info->sinfo_assoc_id);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
if (s_info->sinfo_stream <= stcb->asoc.streamoutcnt) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
memcpy(&stcb->asoc.def_send, s_info, min(optsize, sizeof(stcb->asoc.def_send)));
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = EINVAL;
}
SCTP_TCB_UNLOCK(stcb);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
SCTP_INP_WLOCK(inp);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
memcpy(&inp->def_send, s_info, min(optsize, sizeof(inp->def_send)));
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_WUNLOCK(inp);
}
}
break;
case SCTP_PEER_ADDR_PARAMS:
/* Applys to the specific association */
{
struct sctp_paddrparams *paddrp;
struct sctp_nets *net;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(paddrp, optval, struct sctp_paddrparams, optsize);
SCTP_FIND_STCB(inp, stcb, paddrp->spp_assoc_id);
net = NULL;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (stcb) {
net = sctp_findnet(stcb, (struct sockaddr *)&paddrp->spp_address);
} else {
/*
* We increment here since
* sctp_findassociation_ep_addr() wil do a
* decrement if it finds the stcb as long as
* the locked tcb (last argument) is NOT a
* TCB.. aka NULL.
*/
SCTP_INP_INCR_REF(inp);
stcb = sctp_findassociation_ep_addr(&inp,
(struct sockaddr *)&paddrp->spp_address,
&net, NULL, NULL);
if (stcb == NULL) {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_INP_DECR_REF(inp);
}
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (stcb && (net == NULL)) {
struct sockaddr *sa;
sa = (struct sockaddr *)&paddrp->spp_address;
#ifdef INET
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (sa->sa_family == AF_INET) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
struct sockaddr_in *sin;
sin = (struct sockaddr_in *)sa;
if (sin->sin_addr.s_addr) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
SCTP_TCB_UNLOCK(stcb);
error = EINVAL;
break;
}
} else
#endif
#ifdef INET6
if (sa->sa_family == AF_INET6) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
struct sockaddr_in6 *sin6;
sin6 = (struct sockaddr_in6 *)sa;
if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
SCTP_TCB_UNLOCK(stcb);
error = EINVAL;
break;
}
} else
#endif
{
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
error = EAFNOSUPPORT;
SCTP_TCB_UNLOCK(stcb);
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
break;
}
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
/* sanity checks */
if ((paddrp->spp_flags & SPP_HB_ENABLE) && (paddrp->spp_flags & SPP_HB_DISABLE)) {
if (stcb)
SCTP_TCB_UNLOCK(stcb);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
return (EINVAL);
}
if ((paddrp->spp_flags & SPP_PMTUD_ENABLE) && (paddrp->spp_flags & SPP_PMTUD_DISABLE)) {
if (stcb)
SCTP_TCB_UNLOCK(stcb);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
return (EINVAL);
}
if (stcb) {
/************************TCB SPECIFIC SET ******************/
/*
* do we change the timer for HB, we run
* only one?
*/
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
int ovh = 0;
if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
ovh = SCTP_MED_OVERHEAD;
} else {
ovh = SCTP_MED_V4_OVERHEAD;
}
if (paddrp->spp_hbinterval)
stcb->asoc.heart_beat_delay = paddrp->spp_hbinterval;
else if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO)
stcb->asoc.heart_beat_delay = 0;
/* network sets ? */
if (net) {
/************************NET SPECIFIC SET ******************/
if (paddrp->spp_flags & SPP_HB_DEMAND) {
/* on demand HB */
if (sctp_send_hb(stcb, 1, net, SCTP_SO_LOCKED) < 0) {
/* asoc destroyed */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
}
if (paddrp->spp_flags & SPP_HB_DISABLE) {
net->dest_state |= SCTP_ADDR_NOHB;
}
if (paddrp->spp_flags & SPP_HB_ENABLE) {
net->dest_state &= ~SCTP_ADDR_NOHB;
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if ((paddrp->spp_flags & SPP_PMTUD_DISABLE) && (paddrp->spp_pathmtu >= SCTP_SMALLEST_PMTU)) {
if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) {
sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net,
SCTP_FROM_SCTP_USRREQ + SCTP_LOC_10);
}
if (paddrp->spp_pathmtu > SCTP_DEFAULT_MINSEGMENT) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
net->mtu = paddrp->spp_pathmtu + ovh;
if (net->mtu < stcb->asoc.smallest_mtu) {
sctp_pathmtu_adjustment(inp, stcb, net, net->mtu);
}
}
}
if (paddrp->spp_flags & SPP_PMTUD_ENABLE) {
if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) {
sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
}
}
if (paddrp->spp_pathmaxrxt)
net->failure_threshold = paddrp->spp_pathmaxrxt;
#ifdef INET
if (paddrp->spp_flags & SPP_IPV4_TOS) {
if (net->ro._l_addr.sin.sin_family == AF_INET) {
net->tos_flowlabel = paddrp->spp_ipv4_tos & 0x000000fc;
}
}
#endif
#ifdef INET6
if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL) {
if (net->ro._l_addr.sin6.sin6_family == AF_INET6) {
net->tos_flowlabel = paddrp->spp_ipv6_flowlabel;
}
}
#endif
} else {
/************************ASSOC ONLY -- NO NET SPECIFIC SET ******************/
if (paddrp->spp_pathmaxrxt)
stcb->asoc.def_net_failure = paddrp->spp_pathmaxrxt;
if (paddrp->spp_flags & SPP_HB_ENABLE) {
/* Turn back on the timer */
stcb->asoc.hb_is_disabled = 0;
sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net);
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if ((paddrp->spp_flags & SPP_PMTUD_DISABLE) && (paddrp->spp_pathmtu >= SCTP_SMALLEST_PMTU)) {
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) {
sctp_timer_stop(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net,
SCTP_FROM_SCTP_USRREQ + SCTP_LOC_10);
}
if (paddrp->spp_pathmtu > SCTP_DEFAULT_MINSEGMENT) {
net->mtu = paddrp->spp_pathmtu + ovh;
if (net->mtu < stcb->asoc.smallest_mtu) {
sctp_pathmtu_adjustment(inp, stcb, net, net->mtu);
}
}
}
}
if (paddrp->spp_flags & SPP_PMTUD_ENABLE) {
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
if (SCTP_OS_TIMER_PENDING(&net->pmtu_timer.timer)) {
sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, inp, stcb, net);
}
}
}
if (paddrp->spp_flags & SPP_HB_DISABLE) {
int cnt_of_unconf = 0;
struct sctp_nets *lnet;
stcb->asoc.hb_is_disabled = 1;
TAILQ_FOREACH(lnet, &stcb->asoc.nets, sctp_next) {
if (lnet->dest_state & SCTP_ADDR_UNCONFIRMED) {
cnt_of_unconf++;
}
}
/*
* stop the timer ONLY if we
* have no unconfirmed
* addresses
*/
if (cnt_of_unconf == 0) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
sctp_timer_stop(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net,
SCTP_FROM_SCTP_USRREQ + SCTP_LOC_11);
}
}
}
if (paddrp->spp_flags & SPP_HB_ENABLE) {
/* start up the timer. */
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
sctp_timer_start(SCTP_TIMER_TYPE_HEARTBEAT, inp, stcb, net);
}
}
#ifdef INET
if (paddrp->spp_flags & SPP_IPV4_TOS)
stcb->asoc.default_tos = paddrp->spp_ipv4_tos & 0x000000fc;
#endif
#ifdef INET6
if (paddrp->spp_flags & SPP_IPV6_FLOWLABEL)
stcb->asoc.default_flowlabel = paddrp->spp_ipv6_flowlabel;
#endif
}
SCTP_TCB_UNLOCK(stcb);
} else {
/************************NO TCB, SET TO default stuff ******************/
SCTP_INP_WLOCK(inp);
/*
* For the TOS/FLOWLABEL stuff you set it
* with the options on the socket
*/
if (paddrp->spp_pathmaxrxt) {
inp->sctp_ep.def_net_failure = paddrp->spp_pathmaxrxt;
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (paddrp->spp_flags & SPP_HB_TIME_IS_ZERO)
inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = 0;
else if (paddrp->spp_hbinterval) {
if (paddrp->spp_hbinterval > SCTP_MAX_HB_INTERVAL)
paddrp->spp_hbinterval = SCTP_MAX_HB_INTERVAL;
inp->sctp_ep.sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(paddrp->spp_hbinterval);
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (paddrp->spp_flags & SPP_HB_ENABLE) {
sctp_feature_off(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
} else if (paddrp->spp_flags & SPP_HB_DISABLE) {
sctp_feature_on(inp, SCTP_PCB_FLAGS_DONOT_HEARTBEAT);
}
SCTP_INP_WUNLOCK(inp);
}
}
break;
case SCTP_RTOINFO:
{
struct sctp_rtoinfo *srto;
uint32_t new_init, new_min, new_max;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(srto, optval, struct sctp_rtoinfo, optsize);
SCTP_FIND_STCB(inp, stcb, srto->srto_assoc_id);
if (stcb) {
if (srto->srto_initial)
new_init = srto->srto_initial;
else
new_init = stcb->asoc.initial_rto;
if (srto->srto_max)
new_max = srto->srto_max;
else
new_max = stcb->asoc.maxrto;
if (srto->srto_min)
new_min = srto->srto_min;
else
new_min = stcb->asoc.minrto;
if ((new_min <= new_init) && (new_init <= new_max)) {
stcb->asoc.initial_rto = new_init;
stcb->asoc.maxrto = new_max;
stcb->asoc.minrto = new_min;
} else {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_WLOCK(inp);
if (srto->srto_initial)
new_init = srto->srto_initial;
else
new_init = inp->sctp_ep.initial_rto;
if (srto->srto_max)
new_max = srto->srto_max;
else
new_max = inp->sctp_ep.sctp_maxrto;
if (srto->srto_min)
new_min = srto->srto_min;
else
new_min = inp->sctp_ep.sctp_minrto;
if ((new_min <= new_init) && (new_init <= new_max)) {
inp->sctp_ep.initial_rto = new_init;
inp->sctp_ep.sctp_maxrto = new_max;
inp->sctp_ep.sctp_minrto = new_min;
} else {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
SCTP_INP_WUNLOCK(inp);
}
}
break;
case SCTP_ASSOCINFO:
{
struct sctp_assocparams *sasoc;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sasoc, optval, struct sctp_assocparams, optsize);
SCTP_FIND_STCB(inp, stcb, sasoc->sasoc_assoc_id);
if (sasoc->sasoc_cookie_life) {
/* boundary check the cookie life */
if (sasoc->sasoc_cookie_life < 1000)
sasoc->sasoc_cookie_life = 1000;
if (sasoc->sasoc_cookie_life > SCTP_MAX_COOKIE_LIFE) {
sasoc->sasoc_cookie_life = SCTP_MAX_COOKIE_LIFE;
}
}
if (stcb) {
if (sasoc->sasoc_asocmaxrxt)
stcb->asoc.max_send_times = sasoc->sasoc_asocmaxrxt;
sasoc->sasoc_number_peer_destinations = stcb->asoc.numnets;
sasoc->sasoc_peer_rwnd = 0;
sasoc->sasoc_local_rwnd = 0;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (sasoc->sasoc_cookie_life) {
- take out a needless panic under invariants for sctp_output.c - Fix addrs's error checking of sctp_sendx(3) when addrcnt is less than SCTP_SMALL_IOVEC_SIZE - re-add back inpcb_bind local address check bypass capability - Fix it so sctp_opt_info is independant of assoc_id postion. - Fix cookie life set to use MSEC_TO_TICKS() macro. - asconf changes o More comment changes/clarifications related to the old local address "not" list which is now an explicit restricted list. o Rename some functions for clarity: - sctp_add/del_local_addr_assoc to xxx_local_addr_restricted() - asconf related iterator functions to sctp_asconf_iterator_xxx() o Fix bug when the same address is deleted and added (and removed from the asconf queue) where the ifa is "freed" twice refcount wise, possibly freeing it completely. o Fix bug in output where the first ASCONF would not go out after the last address is changed (e.g. only goes out when retransmitted). o Fix bug where multiple ASCONFs can be bundled in the same packet with the and with the same serial numbers. o Fix asconf stcb iterator to not send ASCONF until after all work queue entries have been processed. o Change behavior so that when the last address is deleted (auto asconf on a bound all endpoint) no action is taken until an address is added; at that time, an ASCONF add+delete is sent (if the assoc is still up). o Fix local address counting so that address scoping is taken into account. o #ifdef SCTP_TIMER_BASED_ASCONF the old timer triggered sending of ASCONF (after an RTO). The default now is to send ASCONF immediately (except for the case of changing/deleting the last usable address). Approved by: re(ken smith)@freebsd.org
2007-07-24 20:06:02 +00:00
stcb->asoc.cookie_life = MSEC_TO_TICKS(sasoc->sasoc_cookie_life);
}
SCTP_TCB_UNLOCK(stcb);
} else {
SCTP_INP_WLOCK(inp);
if (sasoc->sasoc_asocmaxrxt)
inp->sctp_ep.max_send_times = sasoc->sasoc_asocmaxrxt;
sasoc->sasoc_number_peer_destinations = 0;
sasoc->sasoc_peer_rwnd = 0;
sasoc->sasoc_local_rwnd = 0;
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (sasoc->sasoc_cookie_life) {
inp->sctp_ep.def_cookie_life = MSEC_TO_TICKS(sasoc->sasoc_cookie_life);
}
SCTP_INP_WUNLOCK(inp);
}
}
break;
case SCTP_INITMSG:
{
struct sctp_initmsg *sinit;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sinit, optval, struct sctp_initmsg, optsize);
SCTP_INP_WLOCK(inp);
if (sinit->sinit_num_ostreams)
inp->sctp_ep.pre_open_stream_count = sinit->sinit_num_ostreams;
if (sinit->sinit_max_instreams)
inp->sctp_ep.max_open_streams_intome = sinit->sinit_max_instreams;
if (sinit->sinit_max_attempts)
inp->sctp_ep.max_init_times = sinit->sinit_max_attempts;
if (sinit->sinit_max_init_timeo)
inp->sctp_ep.initial_init_rto_max = sinit->sinit_max_init_timeo;
SCTP_INP_WUNLOCK(inp);
}
break;
case SCTP_PRIMARY_ADDR:
{
struct sctp_setprim *spa;
struct sctp_nets *net, *lnet;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(spa, optval, struct sctp_setprim, optsize);
SCTP_FIND_STCB(inp, stcb, spa->ssp_assoc_id);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
net = NULL;
if (stcb) {
net = sctp_findnet(stcb, (struct sockaddr *)&spa->ssp_addr);
} else {
/*
* We increment here since
* sctp_findassociation_ep_addr() wil do a
* decrement if it finds the stcb as long as
* the locked tcb (last argument) is NOT a
* TCB.. aka NULL.
*/
SCTP_INP_INCR_REF(inp);
stcb = sctp_findassociation_ep_addr(&inp,
(struct sockaddr *)&spa->ssp_addr,
&net, NULL, NULL);
if (stcb == NULL) {
SCTP_INP_DECR_REF(inp);
}
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if ((stcb) && (net)) {
if ((net != stcb->asoc.primary_destination) &&
(!(net->dest_state & SCTP_ADDR_UNCONFIRMED))) {
/* Ok we need to set it */
lnet = stcb->asoc.primary_destination;
if (sctp_set_primary_addr(stcb, (struct sockaddr *)NULL, net) == 0) {
if (net->dest_state & SCTP_ADDR_SWITCH_PRIMARY) {
net->dest_state |= SCTP_ADDR_DOUBLE_SWITCH;
}
net->dest_state |= SCTP_ADDR_SWITCH_PRIMARY;
}
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = EINVAL;
}
if (stcb) {
SCTP_TCB_UNLOCK(stcb);
}
}
break;
case SCTP_SET_DYNAMIC_PRIMARY:
{
union sctp_sockstore *ss;
error = priv_check(curthread,
PRIV_NETINET_RESERVEDPORT);
if (error)
break;
SCTP_CHECK_AND_CAST(ss, optval, union sctp_sockstore, optsize);
/* SUPER USER CHECK? */
error = sctp_dynamic_set_primary(&ss->sa, vrf_id);
}
break;
case SCTP_SET_PEER_PRIMARY_ADDR:
{
struct sctp_setpeerprim *sspp;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(sspp, optval, struct sctp_setpeerprim, optsize);
SCTP_FIND_STCB(inp, stcb, sspp->sspp_assoc_id);
if (stcb != NULL) {
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
struct sctp_ifa *ifa;
ifa = sctp_find_ifa_by_addr((struct sockaddr *)&sspp->sspp_addr,
- send call has a reference to uio->uio_resid in the recent send code, but uio may be NULL on sendfile calls. Change to use sndlen variable. - EMSGSIZE is not being returned in non-blocking mode and needs a small tweak to look if the msg would ever fit when returning EWOULDBLOCK. - FWD-TSN has a bug in stream processing which could cause a panic. This is a follow on to the codenomicon fix. - PDAPI level 1 and 2 do not work unless the reader gets his returned buffer full. Fix so we can break out when at level 1 or 2. - Fix fast-handoff features to copy across properly on accepted sockets - Fix sctp_peeloff() system call when no true system call exists to screen arguments for errors. In cases where a real system call exists the system call itself does this. - Fix raddr leak in recent add-ip code change for bundled asconfs (even when non-bundled asconfs are received) - Make sure ipi_addr lock is held when walking global addr list. Need to change this lock type to a rwlock(). - Add don't wake flag on both input and output when the socket is closing. - When deleting an address verify the interface is correct before allowing the delete to process. This protects panda and unnumbered. - Clean up old sysctl stuff and get rid of the old Open/Net BSD structures. - Add a function to watch the ranges in the sysctl sets. - When appending in the reassembly queue, validate that the assoc has not gone to about to be freed. If so (in the middle) abort out. Note this especially effects MAC I think due to the lock/unlock they do (or with LOCK testing in place). - Netstat patch to get rid of warnings. - Make sure that no data gets queued to inactive/unconfirmed destinations. This especially effect CMT but also makes a impact on regular SCTP as well. - During init collision when we detect seq number out of sync we need to treat it like Case C and discard the cookie (no invarient needed here). - Atomic access to the random store. - When we declare a vtag good, we need to shove it into the time wait hash to prevent further use. When the tag is put into the assoc hash, we need to remove it from the twait hash (where it will surely be). This prevents duplicate tag assignments. - Move decr-ref count to better protect sysctl out of data. - ltrace error corrections in sctp6_usrreq.c - Add hook for interface up/down to be sent to us. - Make sysctl() exported structures independent of processor architecture. - Fix route and src addr cache clearing for delete address case. - Make sure address marked SCTP_DEL_IP_ADDRESS is never selected as src addr. - in icmp handling fixed so we actually look at the icmp codes to figure out what to do. - Modified mobility code. Reception of DELETE IP ADDRESS for a primary destination and SET PRIMARY for a new primary destination is used for retransmission trigger to the new primary destination. Also, in this case, destination of chunks in send_queue are changed to the new primary destination. - Fix so that we disallow sending by mbuf to ever have EEOR mode set upon it. Approved by: re@freebsd.org (B Mah)
2007-09-08 17:48:46 +00:00
stcb->asoc.vrf_id, SCTP_ADDR_NOT_LOCKED);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
if (ifa == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
error = EINVAL;
goto out_of_it;
}
if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
/*
* Must validate the ifa found is in
* our ep
*/
struct sctp_laddr *laddr;
int found = 0;
LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
if (laddr->ifa == NULL) {
SCTPDBG(SCTP_DEBUG_OUTPUT1, "%s: NULL ifa\n",
__FUNCTION__);
continue;
}
if (laddr->ifa == ifa) {
found = 1;
break;
}
}
if (!found) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
error = EINVAL;
goto out_of_it;
}
}
if (sctp_set_primary_ip_address_sa(stcb,
(struct sockaddr *)&sspp->sspp_addr) != 0) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = EINVAL;
}
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
out_of_it:
SCTP_TCB_UNLOCK(stcb);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
}
break;
case SCTP_BINDX_ADD_ADDR:
{
struct sctp_getaddresses *addrs;
size_t sz;
struct thread *td;
td = (struct thread *)p;
SCTP_CHECK_AND_CAST(addrs, optval, struct sctp_getaddresses,
optsize);
#ifdef INET
if (addrs->addr->sa_family == AF_INET) {
sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in);
if (optsize < sz) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
if (td != NULL && (error = prison_local_ip4(td->td_ucred, &(((struct sockaddr_in *)(addrs->addr))->sin_addr)))) {
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error);
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
break;
}
} else
#endif
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
#ifdef INET6
if (addrs->addr->sa_family == AF_INET6) {
sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in6);
if (optsize < sz) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
if (td != NULL && (error = prison_local_ip6(td->td_ucred, &(((struct sockaddr_in6 *)(addrs->addr))->sin6_addr),
(SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error);
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
break;
}
} else
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
#endif
{
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
error = EAFNOSUPPORT;
break;
}
sctp_bindx_add_address(so, inp, addrs->addr,
addrs->sget_assoc_id, vrf_id,
&error, p);
}
break;
case SCTP_BINDX_REM_ADDR:
{
struct sctp_getaddresses *addrs;
size_t sz;
struct thread *td;
td = (struct thread *)p;
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SCTP_CHECK_AND_CAST(addrs, optval, struct sctp_getaddresses, optsize);
#ifdef INET
if (addrs->addr->sa_family == AF_INET) {
sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in);
if (optsize < sz) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
if (td != NULL && (error = prison_local_ip4(td->td_ucred, &(((struct sockaddr_in *)(addrs->addr))->sin_addr)))) {
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error);
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
break;
}
} else
#endif
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
#ifdef INET6
if (addrs->addr->sa_family == AF_INET6) {
sz = sizeof(struct sctp_getaddresses) - sizeof(struct sockaddr) + sizeof(struct sockaddr_in6);
if (optsize < sz) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
break;
}
if (td != NULL && (error = prison_local_ip6(td->td_ucred, &(((struct sockaddr_in6 *)(addrs->addr))->sin6_addr),
(SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, error);
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
break;
}
} else
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
#endif
{
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
error = EAFNOSUPPORT;
break;
}
sctp_bindx_delete_address(so, inp, addrs->addr,
addrs->sget_assoc_id, vrf_id,
&error);
}
break;
default:
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOPROTOOPT);
error = ENOPROTOOPT;
break;
} /* end switch (opt) */
return (error);
}
int
sctp_ctloutput(struct socket *so, struct sockopt *sopt)
{
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
void *optval = NULL;
size_t optsize = 0;
struct sctp_inpcb *inp;
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
void *p;
int error = 0;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0) {
/* I made the same as TCP since we are not setup? */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (ECONNRESET);
}
if (sopt->sopt_level != IPPROTO_SCTP) {
/* wrong proto level... send back up to IP */
#ifdef INET6
if (INP_CHECK_SOCKAF(so, AF_INET6))
error = ip6_ctloutput(so, sopt);
#endif /* INET6 */
#if defined(INET) && defined (INET6)
else
#endif
#ifdef INET
error = ip_ctloutput(so, sopt);
#endif
return (error);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
optsize = sopt->sopt_valsize;
if (optsize) {
SCTP_MALLOC(optval, void *, optsize, SCTP_M_SOCKOPT);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if (optval == NULL) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOBUFS);
return (ENOBUFS);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = sooptcopyin(sopt, optval, optsize, optsize);
if (error) {
SCTP_FREE(optval, SCTP_M_SOCKOPT);
goto out;
}
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
p = (void *)sopt->sopt_td;
if (sopt->sopt_dir == SOPT_SET) {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = sctp_setopt(so, sopt->sopt_name, optval, optsize, p);
} else if (sopt->sopt_dir == SOPT_GET) {
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
error = sctp_getopt(so, sopt->sopt_name, optval, &optsize, p);
} else {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
if ((error == 0) && (optval != NULL)) {
error = sooptcopyout(sopt, optval, optsize);
SCTP_FREE(optval, SCTP_M_SOCKOPT);
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
} else if (optval != NULL) {
SCTP_FREE(optval, SCTP_M_SOCKOPT);
}
out:
return (error);
}
#ifdef INET
static int
sctp_connect(struct socket *so, struct sockaddr *addr, struct thread *p)
{
int error = 0;
int create_lock_on = 0;
uint32_t vrf_id;
struct sctp_inpcb *inp;
struct sctp_tcb *stcb = NULL;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0) {
/* I made the same as TCP since we are not setup? */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (ECONNRESET);
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
if (addr == NULL) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
return EINVAL;
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
}
switch (addr->sa_family) {
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
#ifdef INET6
case AF_INET6:
{
struct sockaddr_in6 *sin6p;
if (addr->sa_len != sizeof(struct sockaddr_in6)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (EINVAL);
}
sin6p = (struct sockaddr_in6 *)addr;
if (p != NULL && (error = prison_remote_ip6(p->td_ucred, &sin6p->sin6_addr)) != 0) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
return (error);
}
break;
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
}
#endif
#ifdef INET
case AF_INET:
{
struct sockaddr_in *sinp;
if (addr->sa_len != sizeof(struct sockaddr_in)) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (EINVAL);
}
sinp = (struct sockaddr_in *)addr;
if (p != NULL && (error = prison_remote_ip4(p->td_ucred, &sinp->sin_addr)) != 0) {
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, error);
return (error);
}
break;
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
}
#endif
default:
MFp4: Bring in updated jail support from bz_jail branch. This enhances the current jail implementation to permit multiple addresses per jail. In addtion to IPv4, IPv6 is supported as well. Due to updated checks it is even possible to have jails without an IP address at all, which basically gives one a chroot with restricted process view, no networking,.. SCTP support was updated and supports IPv6 in jails as well. Cpuset support permits jails to be bound to specific processor sets after creation. Jails can have an unrestricted (no duplicate protection, etc.) name in addition to the hostname. The jail name cannot be changed from within a jail and is considered to be used for management purposes or as audit-token in the future. DDB 'show jails' command was added to aid debugging. Proper compat support permits 32bit jail binaries to be used on 64bit systems to manage jails. Also backward compatibility was preserved where possible: for jail v1 syscalls, as well as with user space management utilities. Both jail as well as prison version were updated for the new features. A gap was intentionally left as the intermediate versions had been used by various patches floating around the last years. Bump __FreeBSD_version for the afore mentioned and in kernel changes. Special thanks to: - Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches and Olivier Houchard (cognet) for initial single-IPv6 patches. - Jeff Roberson (jeff) and Randall Stewart (rrs) for their help, ideas and review on cpuset and SCTP support. - Robert Watson (rwatson) for lots and lots of help, discussions, suggestions and review of most of the patch at various stages. - John Baldwin (jhb) for his help. - Simon L. Nielsen (simon) as early adopter testing changes on cluster machines as well as all the testers and people who provided feedback the last months on freebsd-jail and other channels. - My employer, CK Software GmbH, for the support so I could work on this. Reviewed by: (see above) MFC after: 3 months (this is just so that I get the mail) X-MFC Before: 7.2-RELEASE if possible
2008-11-29 14:32:14 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EAFNOSUPPORT);
return (EAFNOSUPPORT);
- fixed autclose to not allow setting on 1-2-1 model. - bounded cookie-life to 1 second minimum in socket option set. - Delayed_ack_time becomes delayed_ack per new socket api document. - Improve port number selection, we now use low/high bounds and no chance of a endless loop. Only one call to random per bind as well. - fixes so set_peer_primary pre-screens addresses to be valid to this host. - maxseg did not allow setting on an assoc basis. We needed to thus track and use an association value instead of a inp value. - Fixed ep get of HB status to report back properly. - use settings flag to tell if assoc level hb is on off not the timer.. since the timer may still run if unconf address are present. - check for crazy ENABLE/DISABLE conditions. - set and get of pmtud (fixed path mtu) not always taking into account ovh. - Getting PMTU info on stcb only needs to return PMTUD_ENABLED if any net is doing PMTU discovery. - Panic or warning fixed to not do so when a valid ip frag is taking place. - sndrcvinfo appearing in both inp and stcb was full size, instead of the non-pad version. This saves about 92 bytes from each struct by carefully converting to use the smaller version. - one-2-one model get(maxseg) would always get ep value, never the tcb's value. - The delayed ack time could be under a tick, this fixes so it bounds it to at least 1 tick for platforms whos tick is more than a ms. - Fragment interleave level set to wrong default value. - Fragment interleave could not set level 0. - Defered stream reset was broken due to a guard check and ntohl issue. - Found two lock order reversals and fixed. - Tighten up address checking, if the user gives an address the sa_len had better be set properly. - Get asoc by assoc-id would return a locked tcb when it was asked not to if the tcb was in the restart hash. - sysctl to dig down and get more association details Reviewed by: gnn
2007-05-28 11:17:24 +00:00
}
SCTP_INP_INCR_REF(inp);
SCTP_ASOC_CREATE_LOCK(inp);
create_lock_on = 1;
if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
(inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
/* Should I really unlock ? */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EFAULT);
error = EFAULT;
goto out_now;
}
#ifdef INET6
if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) &&
(addr->sa_family == AF_INET6)) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
goto out_now;
}
#endif /* INET6 */
if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) ==
SCTP_PCB_FLAGS_UNBOUND) {
/* Bind a ephemeral port */
- take out a needless panic under invariants for sctp_output.c - Fix addrs's error checking of sctp_sendx(3) when addrcnt is less than SCTP_SMALL_IOVEC_SIZE - re-add back inpcb_bind local address check bypass capability - Fix it so sctp_opt_info is independant of assoc_id postion. - Fix cookie life set to use MSEC_TO_TICKS() macro. - asconf changes o More comment changes/clarifications related to the old local address "not" list which is now an explicit restricted list. o Rename some functions for clarity: - sctp_add/del_local_addr_assoc to xxx_local_addr_restricted() - asconf related iterator functions to sctp_asconf_iterator_xxx() o Fix bug when the same address is deleted and added (and removed from the asconf queue) where the ifa is "freed" twice refcount wise, possibly freeing it completely. o Fix bug in output where the first ASCONF would not go out after the last address is changed (e.g. only goes out when retransmitted). o Fix bug where multiple ASCONFs can be bundled in the same packet with the and with the same serial numbers. o Fix asconf stcb iterator to not send ASCONF until after all work queue entries have been processed. o Change behavior so that when the last address is deleted (auto asconf on a bound all endpoint) no action is taken until an address is added; at that time, an ASCONF add+delete is sent (if the assoc is still up). o Fix local address counting so that address scoping is taken into account. o #ifdef SCTP_TIMER_BASED_ASCONF the old timer triggered sending of ASCONF (after an RTO). The default now is to send ASCONF immediately (except for the case of changing/deleting the last usable address). Approved by: re(ken smith)@freebsd.org
2007-07-24 20:06:02 +00:00
error = sctp_inpcb_bind(so, NULL, NULL, p);
if (error) {
goto out_now;
}
}
/* Now do we connect? */
if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
(sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE))) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
error = EINVAL;
goto out_now;
}
if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) &&
(inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) {
/* We are already connected AND the TCP model */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE);
error = EADDRINUSE;
goto out_now;
}
if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
SCTP_INP_RLOCK(inp);
stcb = LIST_FIRST(&inp->sctp_asoc_list);
SCTP_INP_RUNLOCK(inp);
} else {
/*
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
* We increment here since sctp_findassociation_ep_addr()
* will do a decrement if it finds the stcb as long as the
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
* locked tcb (last argument) is NOT a TCB.. aka NULL.
*/
SCTP_INP_INCR_REF(inp);
stcb = sctp_findassociation_ep_addr(&inp, addr, NULL, NULL, NULL);
if (stcb == NULL) {
SCTP_INP_DECR_REF(inp);
} else {
SCTP_TCB_UNLOCK(stcb);
}
}
if (stcb != NULL) {
/* Already have or am bring up an association */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EALREADY);
error = EALREADY;
goto out_now;
}
vrf_id = inp->def_vrf_id;
/* We are GOOD to go */
stcb = sctp_aloc_assoc(inp, addr, &error, 0, vrf_id, p);
if (stcb == NULL) {
/* Gak! no memory */
goto out_now;
}
if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
stcb->sctp_ep->sctp_flags |= SCTP_PCB_FLAGS_CONNECTED;
/* Set the connected flag so we can queue data */
SOCKBUF_LOCK(&so->so_rcv);
so->so_rcv.sb_state &= ~SBS_CANTRCVMORE;
SOCKBUF_UNLOCK(&so->so_rcv);
SOCKBUF_LOCK(&so->so_snd);
so->so_snd.sb_state &= ~SBS_CANTSENDMORE;
SOCKBUF_UNLOCK(&so->so_snd);
SOCK_LOCK(so);
so->so_state &= ~SS_ISDISCONNECTING;
SOCK_UNLOCK(so);
soisconnecting(so);
}
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_SET_STATE(&stcb->asoc, SCTP_STATE_COOKIE_WAIT);
(void)SCTP_GETTIME_TIMEVAL(&stcb->asoc.time_entered);
/* initialize authentication parameters for the assoc */
sctp_initialize_auth_params(inp, stcb);
sctp_send_initiate(inp, stcb, SCTP_SO_LOCKED);
SCTP_TCB_UNLOCK(stcb);
out_now:
if (create_lock_on) {
SCTP_ASOC_CREATE_UNLOCK(inp);
}
SCTP_INP_DECR_REF(inp);
return error;
}
#endif
int
sctp_listen(struct socket *so, int backlog, struct thread *p)
{
/*
* Note this module depends on the protocol processing being called
* AFTER any socket level flags and backlog are applied to the
* socket. The traditional way that the socket flags are applied is
* AFTER protocol processing. We have made a change to the
* sys/kern/uipc_socket.c module to reverse this but this MUST be in
* place if the socket API for SCTP is to work properly.
*/
int error = 0;
struct sctp_inpcb *inp;
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0) {
/* I made the same as TCP since we are not setup? */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (ECONNRESET);
}
if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
/* See if we have a listener */
struct sctp_inpcb *tinp;
union sctp_sockstore store, *sp;
sp = &store;
if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
/* not bound all */
struct sctp_laddr *laddr;
LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
memcpy(&store, &laddr->ifa->address, sizeof(store));
switch (sp->sa.sa_family) {
#ifdef INET
case AF_INET:
sp->sin.sin_port = inp->sctp_lport;
break;
#endif
#ifdef INET6
case AF_INET6:
sp->sin6.sin6_port = inp->sctp_lport;
break;
#endif
default:
break;
}
tinp = sctp_pcb_findep(&sp->sa, 0, 0, inp->def_vrf_id);
if (tinp && (tinp != inp) &&
((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) &&
((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) &&
(tinp->sctp_socket->so_qlimit)) {
/*
* we have a listener already and
* its not this inp.
*/
SCTP_INP_DECR_REF(tinp);
return (EADDRINUSE);
} else if (tinp) {
SCTP_INP_DECR_REF(tinp);
}
}
} else {
/* Setup a local addr bound all */
memset(&store, 0, sizeof(store));
switch (sp->sa.sa_family) {
#ifdef INET
case AF_INET:
store.sin.sin_port = inp->sctp_lport;
break;
#endif
#ifdef INET6
case AF_INET6:
sp->sin6.sin6_port = inp->sctp_lport;
break;
#endif
default:
break;
}
#ifdef INET6
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
store.sa.sa_family = AF_INET6;
store.sa.sa_len = sizeof(struct sockaddr_in6);
}
#endif
#ifdef INET
if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
store.sa.sa_family = AF_INET;
store.sa.sa_len = sizeof(struct sockaddr_in);
}
#endif
tinp = sctp_pcb_findep(&sp->sa, 0, 0, inp->def_vrf_id);
if (tinp && (tinp != inp) &&
((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) == 0) &&
((tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) == 0) &&
(tinp->sctp_socket->so_qlimit)) {
/*
* we have a listener already and its not
* this inp.
*/
SCTP_INP_DECR_REF(tinp);
return (EADDRINUSE);
} else if (tinp) {
SCTP_INP_DECR_REF(inp);
}
}
}
SCTP_INP_RLOCK(inp);
#ifdef SCTP_LOCK_LOGGING
if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_LOCK_LOGGING_ENABLE) {
sctp_log_lock(inp, (struct sctp_tcb *)NULL, SCTP_LOG_LOCK_SOCK);
}
#endif
SOCK_LOCK(so);
error = solisten_proto_check(so);
if (error) {
SOCK_UNLOCK(so);
SCTP_INP_RUNLOCK(inp);
return (error);
}
if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
(inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
/*
* The unlucky case - We are in the tcp pool with this guy.
* - Someone else is in the main inp slot. - We must move
* this guy (the listener) to the main slot - We must then
* move the guy that was listener to the TCP Pool.
*/
if (sctp_swap_inpcb_for_listen(inp)) {
goto in_use;
}
}
if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) &&
(inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED)) {
/* We are already connected AND the TCP model */
in_use:
SCTP_INP_RUNLOCK(inp);
SOCK_UNLOCK(so);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EADDRINUSE);
return (EADDRINUSE);
}
SCTP_INP_RUNLOCK(inp);
if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
/* We must do a bind. */
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SOCK_UNLOCK(so);
- take out a needless panic under invariants for sctp_output.c - Fix addrs's error checking of sctp_sendx(3) when addrcnt is less than SCTP_SMALL_IOVEC_SIZE - re-add back inpcb_bind local address check bypass capability - Fix it so sctp_opt_info is independant of assoc_id postion. - Fix cookie life set to use MSEC_TO_TICKS() macro. - asconf changes o More comment changes/clarifications related to the old local address "not" list which is now an explicit restricted list. o Rename some functions for clarity: - sctp_add/del_local_addr_assoc to xxx_local_addr_restricted() - asconf related iterator functions to sctp_asconf_iterator_xxx() o Fix bug when the same address is deleted and added (and removed from the asconf queue) where the ifa is "freed" twice refcount wise, possibly freeing it completely. o Fix bug in output where the first ASCONF would not go out after the last address is changed (e.g. only goes out when retransmitted). o Fix bug where multiple ASCONFs can be bundled in the same packet with the and with the same serial numbers. o Fix asconf stcb iterator to not send ASCONF until after all work queue entries have been processed. o Change behavior so that when the last address is deleted (auto asconf on a bound all endpoint) no action is taken until an address is added; at that time, an ASCONF add+delete is sent (if the assoc is still up). o Fix local address counting so that address scoping is taken into account. o #ifdef SCTP_TIMER_BASED_ASCONF the old timer triggered sending of ASCONF (after an RTO). The default now is to send ASCONF immediately (except for the case of changing/deleting the last usable address). Approved by: re(ken smith)@freebsd.org
2007-07-24 20:06:02 +00:00
if ((error = sctp_inpcb_bind(so, NULL, NULL, p))) {
/* bind error, probably perm */
return (error);
}
- Copyright updates (aka 2007) - ZONE get now also take a type cast so it does the cast like mtod does. - New macro SCTP_LIST_EMPTY, which in bsd is just LIST_EMPTY - Removal of const in some of the static hmac functions (not needed) - Store length changes to allow for new fields in auth - Auth code updated to current draft (this should be the RFC version we think). - use uint8_t instead of u_char in LOOPBACK address comparison - Some u_int32_t converted to uint32_t (in crc code) - A bug was found in the mib counts for ordered/unordered count, this was fixed (was referencing a freed mbuf). - SCTP_ASOCLOG_OF_TSNS added (code will probably disappear after my testing completes. It allows us to keep a small log on each assoc of the last 40 TSN's in/out and stream assignment. It is NOT in options and so is only good for private builds. - Some CMT changes in prep for Jana fixing his problem with reneging when CMT is enabled (Concurrent Multipath Transfer = CMT). - Some missing mib stats added. - Correction to number of open assoc's count in mib - Correction to os_bsd.h to get right sha2 macros - Add of special AUTH_04 flags so you can compile the code with the old format (in case the peer does not yet support the latest auth code). - Nonce sum was incorrectly being set in when ecn_nonce was NOT on. - LOR in listen with implicit bind found and fixed. - Moved away from using mbuf's for socket options to using just data pointers. The mbufs were used to harmonize NetBSD code since both Net and Open used this method. We have decided to move away from that and more conform to FreeBSD style (which makes more sense). - Very very nasty bug found in some of my "debug" code. The cookie_how collision case tracking had an endless loop in it if you got a second retransmission of a cookie collision case. This would lock up a CPU .. ugly.. - auth function goes to using size_t instead of int which conforms to socketapi better - Found the nasty bug that happens after 9 days of testing.. you get the data chunk, deliver it and due to the reference to a ch-> that every now and then has been deleted (depending on the postion in the mbuf) you have an invalid ch->ch.flags.. and thus you don't advance the stream sequence number.. so you block the stream permanently. The fix is to make local variables of these guys and set them up before you have any chance of trimming the mbuf. - style fix in sctp_util.h, not sure how this got bad maybe in the last patch? (aka it may not be in the real source). - Found interesting bug when using the extended snd/rcv info where we would get an error on receiving with this. Thats because it was NOT padded to the same size as the snd_rcv info. We increase (add the pad) so the two structs are the same size in sctp_uio.h - In sctp_usrreq.c one of the most common things we did for socket options was to cast the pointer and validate the size. This as been macro-ized to help make the code more readable. - in sctputil.c two things, the socketapi class found a missing flag type (the next msg is a notification) and a missing scope recovery was also fixed. Reviewed by: gnn
2007-02-12 23:24:31 +00:00
SOCK_LOCK(so);
}
/* It appears for 7.0 and on, we must always call this. */
solisten_proto(so, backlog);
if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) {
/* remove the ACCEPTCONN flag for one-to-many sockets */
so->so_options &= ~SO_ACCEPTCONN;
}
if (backlog == 0) {
/* turning off listen */
so->so_options &= ~SO_ACCEPTCONN;
}
SOCK_UNLOCK(so);
return (error);
}
static int sctp_defered_wakeup_cnt = 0;
int
sctp_accept(struct socket *so, struct sockaddr **addr)
{
struct sctp_tcb *stcb;
struct sctp_inpcb *inp;
union sctp_sockstore store;
#ifdef INET6
int error;
#endif
inp = (struct sctp_inpcb *)so->so_pcb;
if (inp == 0) {
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (ECONNRESET);
}
SCTP_INP_RLOCK(inp);
if (inp->sctp_flags & SCTP_PCB_FLAGS_UDPTYPE) {
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EOPNOTSUPP);
return (EOPNOTSUPP);
}
if (so->so_state & SS_ISDISCONNECTED) {
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ECONNABORTED);
return (ECONNABORTED);
}
stcb = LIST_FIRST(&inp->sctp_asoc_list);
if (stcb == NULL) {
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return (ECONNRESET);
}
SCTP_TCB_LOCK(stcb);
SCTP_INP_RUNLOCK(inp);
store = stcb->asoc.primary_destination->ro._l_addr;
stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
SCTP_TCB_UNLOCK(stcb);
switch (store.sa.sa_family) {
#ifdef INET
case AF_INET:
{
struct sockaddr_in *sin;
SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin);
if (sin == NULL)
return (ENOMEM);
sin->sin_family = AF_INET;
sin->sin_len = sizeof(*sin);
sin->sin_port = ((struct sockaddr_in *)&store)->sin_port;
sin->sin_addr = ((struct sockaddr_in *)&store)->sin_addr;
*addr = (struct sockaddr *)sin;
break;
}
#endif
#ifdef INET6
case AF_INET6:
{
struct sockaddr_in6 *sin6;
SCTP_MALLOC_SONAME(sin6, struct sockaddr_in6 *, sizeof *sin6);
if (sin6 == NULL)
return (ENOMEM);
sin6->sin6_family = AF_INET6;
sin6->sin6_len = sizeof(*sin6);
sin6->sin6_port = ((struct sockaddr_in6 *)&store)->sin6_port;
sin6->sin6_addr = ((struct sockaddr_in6 *)&store)->sin6_addr;
if ((error = sa6_recoverscope(sin6)) != 0) {
SCTP_FREE_SONAME(sin6);
return (error);
}
*addr = (struct sockaddr *)sin6;
break;
}
#endif
default:
/* TSNH */
break;
}
/* Wake any delayed sleep action */
if (inp->sctp_flags & SCTP_PCB_FLAGS_DONT_WAKE) {
SCTP_INP_WLOCK(inp);
inp->sctp_flags &= ~SCTP_PCB_FLAGS_DONT_WAKE;
if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEOUTPUT) {
inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
SCTP_INP_WUNLOCK(inp);
SOCKBUF_LOCK(&inp->sctp_socket->so_snd);
if (sowriteable(inp->sctp_socket)) {
sowwakeup_locked(inp->sctp_socket);
} else {
SOCKBUF_UNLOCK(&inp->sctp_socket->so_snd);
}
SCTP_INP_WLOCK(inp);
}
if (inp->sctp_flags & SCTP_PCB_FLAGS_WAKEINPUT) {
inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
SCTP_INP_WUNLOCK(inp);
SOCKBUF_LOCK(&inp->sctp_socket->so_rcv);
if (soreadable(inp->sctp_socket)) {
sctp_defered_wakeup_cnt++;
sorwakeup_locked(inp->sctp_socket);
} else {
SOCKBUF_UNLOCK(&inp->sctp_socket->so_rcv);
}
SCTP_INP_WLOCK(inp);
}
SCTP_INP_WUNLOCK(inp);
}
if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
SCTP_TCB_LOCK(stcb);
sctp_free_assoc(inp, stcb, SCTP_NORMAL_PROC, SCTP_FROM_SCTP_USRREQ + SCTP_LOC_7);
}
return (0);
}
#ifdef INET
int
sctp_ingetaddr(struct socket *so, struct sockaddr **addr)
{
struct sockaddr_in *sin;
uint32_t vrf_id;
struct sctp_inpcb *inp;
struct sctp_ifa *sctp_ifa;
/*
* Do the malloc first in case it blocks.
*/
SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin);
if (sin == NULL)
return (ENOMEM);
sin->sin_family = AF_INET;
sin->sin_len = sizeof(*sin);
inp = (struct sctp_inpcb *)so->so_pcb;
if (!inp) {
SCTP_FREE_SONAME(sin);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return ECONNRESET;
}
SCTP_INP_RLOCK(inp);
sin->sin_port = inp->sctp_lport;
if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
struct sctp_tcb *stcb;
struct sockaddr_in *sin_a;
struct sctp_nets *net;
int fnd;
stcb = LIST_FIRST(&inp->sctp_asoc_list);
if (stcb == NULL) {
goto notConn;
}
fnd = 0;
sin_a = NULL;
SCTP_TCB_LOCK(stcb);
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
sin_a = (struct sockaddr_in *)&net->ro._l_addr;
if (sin_a == NULL)
/* this will make coverity happy */
continue;
if (sin_a->sin_family == AF_INET) {
fnd = 1;
break;
}
}
if ((!fnd) || (sin_a == NULL)) {
/* punt */
SCTP_TCB_UNLOCK(stcb);
goto notConn;
}
vrf_id = inp->def_vrf_id;
sctp_ifa = sctp_source_address_selection(inp,
stcb,
(sctp_route_t *) & net->ro,
net, 0, vrf_id);
if (sctp_ifa) {
sin->sin_addr = sctp_ifa->address.sin.sin_addr;
sctp_free_ifa(sctp_ifa);
}
SCTP_TCB_UNLOCK(stcb);
} else {
/* For the bound all case you get back 0 */
notConn:
sin->sin_addr.s_addr = 0;
}
} else {
/* Take the first IPv4 address in the list */
struct sctp_laddr *laddr;
int fnd = 0;
LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
if (laddr->ifa->address.sa.sa_family == AF_INET) {
struct sockaddr_in *sin_a;
sin_a = (struct sockaddr_in *)&laddr->ifa->address.sa;
sin->sin_addr = sin_a->sin_addr;
fnd = 1;
break;
}
}
if (!fnd) {
SCTP_FREE_SONAME(sin);
SCTP_INP_RUNLOCK(inp);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
return ENOENT;
}
}
SCTP_INP_RUNLOCK(inp);
(*addr) = (struct sockaddr *)sin;
return (0);
}
int
sctp_peeraddr(struct socket *so, struct sockaddr **addr)
{
struct sockaddr_in *sin = (struct sockaddr_in *)*addr;
int fnd;
struct sockaddr_in *sin_a;
struct sctp_inpcb *inp;
struct sctp_tcb *stcb;
struct sctp_nets *net;
/* Do the malloc first in case it blocks. */
inp = (struct sctp_inpcb *)so->so_pcb;
if ((inp == NULL) ||
((inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) == 0)) {
/* UDP type and listeners will drop out here */
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOTCONN);
return (ENOTCONN);
}
SCTP_MALLOC_SONAME(sin, struct sockaddr_in *, sizeof *sin);
if (sin == NULL)
return (ENOMEM);
sin->sin_family = AF_INET;
sin->sin_len = sizeof(*sin);
/* We must recapture incase we blocked */
inp = (struct sctp_inpcb *)so->so_pcb;
if (!inp) {
SCTP_FREE_SONAME(sin);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return ECONNRESET;
}
SCTP_INP_RLOCK(inp);
stcb = LIST_FIRST(&inp->sctp_asoc_list);
if (stcb) {
SCTP_TCB_LOCK(stcb);
}
SCTP_INP_RUNLOCK(inp);
if (stcb == NULL) {
SCTP_FREE_SONAME(sin);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, EINVAL);
return ECONNRESET;
}
fnd = 0;
TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
sin_a = (struct sockaddr_in *)&net->ro._l_addr;
if (sin_a->sin_family == AF_INET) {
fnd = 1;
sin->sin_port = stcb->rport;
sin->sin_addr = sin_a->sin_addr;
break;
}
}
SCTP_TCB_UNLOCK(stcb);
if (!fnd) {
/* No IPv4 address */
SCTP_FREE_SONAME(sin);
- Fix address add handling to clear cached routes and source addresses when peer acks the add in case the routing table changes. - Fix sctp_lower_sosend to send shutdown chunk for mbuf send case when sndlen = 0 and sinfoflag = SCTP_EOF - Fix sctp_lower_sosend for SCTP_ABORT mbuf send case with null data, So that it does not send the "null" data mbuf out and cause it to get freed twice. - Fix so auto-asconf sysctl actually effect the socket's asconf state. - Do not allow SCTP_AUTO_ASCONF option to be used on subset bound sockets. - Memset bug in sctp_output.c (arguments were reversed) submitted found and reported by Dave Jones (davej@codemonkey.org.uk). - PD-API point needs to be invoked >= not just > to conform to socket api draft this fixes sctp_indata.c in the two places need to be >=. - move M_NOTIFICATION to use M_PROTO5. - PEER_ADDR_PARAMS did not fail properly if you specify an address that is not in the association with a valid assoc_id. This meant you got or set the stcb level values instead of the destination you thought you were going to get/set. Now validate if the stcb is non-null and the net is NULL that the sa_family is set and the address is unspecified otherwise return an error. - The thread based iterator could crash if associations were freed at the exact time it was running. rework the worker thread to use the increment/decrement to prevent this and no longer use the markers that the timer based iterator uses. - Fix the memleak in sctp_add_addr_to_vrf() for the case when it is detected that ifa is already pointing to a ifn. - Fix it so that if someone is so insane that they drop the send window below the minimal add mark, they still can send. - Changed all state for associations to use mask safe macro. - During front states in association freeing in sctp_inpcbfree, we had a locking problem where locks were not in place where they should have been. - Free association calls were not testing the return value in sctp_inpcb_free() properly... others should be cast void returns where we don't care about the return value. - If a reference count is held on an assoc, even from the "force free" we should not do the actual free.. but instead let the timer free it. - When we enter sctp_input(), if the SCTP_ASOC_ABOUT_TO_BE_FREED flag is set, we must NOT process the packet but handle it like ootb. This is because while freeing an assoc we release the locks to get all the higher order locks so we can purge all the hash tables. This leaves a hole if a packet comes in just at that point. Now sctp_common_input_processing() will call the ootb code in such a case. - Change MBUF M_NOTIFICATION to use M_PROTO5 (per Sam L). This makes it so we don't have a conflict (I think this is a covertity change). We made this change AFTER some conversation and looking to make sure that M_PROTO5 does not have a problem between SCTP and the 802.11 stuff (which is the only other place its used). - Fixed lock order reversal and missing atomic protection around locked_tcb during association lookup and the 1-2-1 model. - Added debug to source address selection. - V6 output must always do checksum even for loopback. - Remove more locks around inp that are not needed for an atomically added/subtracted ref count. - slight optimization in the way we zero the array in sctp_sack_check() - It was possible to respond to a ABORT() with bad checksum with a PKT-DROP. This lead to a PKT-DROP/ABORT war. Add code to NOT send a PKT-DROP to any ABORT(). - Add an option for local logging (useful for macintosh or when you need better performing during debugging). Note no commands are here to get the log info, you must just use kgdb. - The timer code needs to be aware of if it needs to call sctp_sack_check() to slide the maps and adjust the cum-ack. This is because it may be out of sync cum-ack wise. - Added threshold managment logging. - If the user picked just the right size, that just filled the send window minus one mtu, we would enter a forever loop not copying and at the same time not blocking. Change from < to <= solves this. - Sysctl added to control the fragment interleave level which defaults to 1. - My rwnd control was not being used to control the rwnd properly (we did not add and subtract to it :-() this is now fixed so we handle small messages (1 byte etc) better to bring our rwnd down more slowly. Approved by: re@freebsd.org (Bruce Mah)
2007-08-24 00:53:53 +00:00
SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_USRREQ, ENOENT);
return ENOENT;
}
(*addr) = (struct sockaddr *)sin;
return (0);
}
#ifdef INET
struct pr_usrreqs sctp_usrreqs = {
.pru_abort = sctp_abort,
.pru_accept = sctp_accept,
.pru_attach = sctp_attach,
.pru_bind = sctp_bind,
.pru_connect = sctp_connect,
.pru_control = in_control,
.pru_close = sctp_close,
.pru_detach = sctp_close,
.pru_sopoll = sopoll_generic,
.pru_flush = sctp_flush,
.pru_disconnect = sctp_disconnect,
.pru_listen = sctp_listen,
.pru_peeraddr = sctp_peeraddr,
.pru_send = sctp_sendm,
.pru_shutdown = sctp_shutdown,
.pru_sockaddr = sctp_ingetaddr,
.pru_sosend = sctp_sosend,
.pru_soreceive = sctp_soreceive
};
#endif
#endif