freebsd-skq/sys/netinet/tcp_usrreq.c

2042 lines
49 KiB
C
Raw Normal View History

/*-
1994-05-24 10:09:53 +00:00
* Copyright (c) 1982, 1986, 1988, 1993
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
* The Regents of the University of California.
* Copyright (c) 2006-2007 Robert N. M. Watson
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
* Copyright (c) 2010-2011 Juniper Networks, Inc.
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
* All rights reserved.
1994-05-24 10:09:53 +00:00
*
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
* Portions of this software were developed by Robert N. M. Watson under
* contract to Juniper Networks, Inc.
*
1994-05-24 10:09:53 +00:00
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. 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.
* 4. Neither the name of the University 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 REGENTS 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 REGENTS 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.
*
* From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
1994-05-24 10:09:53 +00:00
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ddb.h"
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#include "opt_inet.h"
#include "opt_inet6.h"
1997-09-16 18:36:06 +00:00
#include "opt_tcpdebug.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/limits.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
1994-05-24 10:09:53 +00:00
#include <sys/mbuf.h>
#ifdef INET6
#include <sys/domain.h>
#endif /* INET6 */
1994-05-24 10:09:53 +00:00
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/protosw.h>
#include <sys/proc.h>
#include <sys/jail.h>
1994-05-24 10:09:53 +00:00
#ifdef DDB
#include <ddb/ddb.h>
#endif
1994-05-24 10:09:53 +00:00
#include <net/if.h>
#include <net/route.h>
#include <net/vnet.h>
1994-05-24 10:09:53 +00:00
This commit marks the first formal contribution of the "Five New TCP Congestion Control Algorithms for FreeBSD" FreeBSD Foundation funded project. More details about the project are available at: http://caia.swin.edu.au/freebsd/5cc/ - Add a KPI and supporting infrastructure to allow modular congestion control algorithms to be used in the net stack. Algorithms can maintain per-connection state if required, and connections maintain their own algorithm pointer, which allows different connections to concurrently use different algorithms. The TCP_CONGESTION socket option can be used with getsockopt()/setsockopt() to programmatically query or change the congestion control algorithm respectively from within an application at runtime. - Integrate the framework with the TCP stack in as least intrusive a manner as possible. Care was also taken to develop the framework in a way that should allow integration with other congestion aware transport protocols (e.g. SCTP) in the future. The hope is that we will one day be able to share a single set of congestion control algorithm modules between all congestion aware transport protocols. - Introduce a new congestion recovery (TF_CONGRECOVERY) state into the TCP stack and use it to decouple the meaning of recovery from a congestion event and recovery from packet loss (TF_FASTRECOVERY) a la RFC2581. ECN and delay based congestion control protocols don't generally need to recover from packet loss and need a different way to note a congestion recovery episode within the stack. - Remove the net.inet.tcp.newreno sysctl, which simplifies some portions of code and ensures the stack always uses the appropriate mechanisms for recovering from packet loss during a congestion recovery episode. - Extract the NewReno congestion control algorithm from the TCP stack and massage it into module form. NewReno is always built into the kernel and will remain the default algorithm for the forseeable future. Implementations of additional different algorithms will become available in the near future. - Bump __FreeBSD_version to 900025 and note in UPDATING that rebuilding code that relies on the size of "struct tcpcb" is required. Many thanks go to the Cisco University Research Program Fund at Community Foundation Silicon Valley and the FreeBSD Foundation. Their support of our work at the Centre for Advanced Internet Architectures, Swinburne University of Technology is greatly appreciated. In collaboration with: David Hayes <dahayes at swin edu au> and Grenville Armitage <garmitage at swin edu au> Sponsored by: Cisco URP, FreeBSD Foundation Reviewed by: rpaulo Tested by: David Hayes (and many others over the years) MFC after: 3 months
2010-11-12 06:41:55 +00:00
#include <netinet/cc.h>
1994-05-24 10:09:53 +00:00
#include <netinet/in.h>
#include <netinet/in_pcb.h>
#include <netinet/in_systm.h>
#include <netinet/in_var.h>
1994-05-24 10:09:53 +00:00
#include <netinet/ip_var.h>
#ifdef INET6
#include <netinet/ip6.h>
#include <netinet6/in6_pcb.h>
#include <netinet6/ip6_var.h>
#include <netinet6/scope6_var.h>
#endif
1994-05-24 10:09:53 +00:00
#include <netinet/tcp_fsm.h>
#include <netinet/tcp_seq.h>
#include <netinet/tcp_timer.h>
#include <netinet/tcp_var.h>
#include <netinet/tcpip.h>
#ifdef TCPDEBUG
1994-05-24 10:09:53 +00:00
#include <netinet/tcp_debug.h>
#endif
#ifdef TCP_OFFLOAD
#include <netinet/tcp_offload.h>
#endif
1994-05-24 10:09:53 +00:00
/*
* TCP protocol interface to socket abstraction.
*/
2004-03-28 15:48:00 +00:00
static int tcp_attach(struct socket *);
#ifdef INET
static int tcp_connect(struct tcpcb *, struct sockaddr *,
struct thread *td);
#endif /* INET */
#ifdef INET6
2002-03-19 21:25:46 +00:00
static int tcp6_connect(struct tcpcb *, struct sockaddr *,
struct thread *td);
#endif /* INET6 */
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
static void tcp_disconnect(struct tcpcb *);
static void tcp_usrclosed(struct tcpcb *);
static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
#ifdef TCPDEBUG
2001-03-12 02:57:42 +00:00
#define TCPDEBUG0 int ostate = 0
#define TCPDEBUG1() ostate = tp ? tp->t_state : 0
#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
tcp_trace(TA_USER, ostate, tp, 0, 0, req)
#else
#define TCPDEBUG0
#define TCPDEBUG1()
#define TCPDEBUG2(req)
#endif
/*
* TCP attaches to socket via pru_attach(), reserving space,
* and an internet control block.
*/
static int
tcp_usr_attach(struct socket *so, int proto, struct thread *td)
{
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
int error;
TCPDEBUG0;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
inp = sotoinpcb(so);
KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
TCPDEBUG1();
2004-03-28 15:48:00 +00:00
error = tcp_attach(so);
if (error)
goto out;
if ((so->so_options & SO_LINGER) && so->so_linger == 0)
so->so_linger = TCP_LINGERTIME;
inp = sotoinpcb(so);
tp = intotcpcb(inp);
out:
TCPDEBUG2(PRU_ATTACH);
return error;
}
/*
* tcp_detach is called when the socket layer loses its final reference
* to the socket, be it a file descriptor reference, a reference from TCP,
* etc. At this point, there is only one case in which we will keep around
* inpcb state: time wait.
*
* This function can probably be re-absorbed back into tcp_usr_detach() now
* that there is a single detach path.
*/
static void
tcp_detach(struct socket *so, struct inpcb *inp)
{
struct tcpcb *tp;
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
INP_WLOCK_ASSERT(inp);
KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
tp = intotcpcb(inp);
if (inp->inp_flags & INP_TIMEWAIT) {
/*
* There are two cases to handle: one in which the time wait
* state is being discarded (INP_DROPPED), and one in which
* this connection will remain in timewait. In the former,
* it is time to discard all state (except tcptw, which has
* already been discarded by the timewait close code, which
* should be further up the call stack somewhere). In the
* latter case, we detach from the socket, but leave the pcb
* present until timewait ends.
*
* XXXRW: Would it be cleaner to free the tcptw here?
*/
if (inp->inp_flags & INP_DROPPED) {
KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
"INP_DROPPED && tp != NULL"));
in_pcbdetach(inp);
in_pcbfree(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
} else {
in_pcbdetach(inp);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
}
} else {
/*
* If the connection is not in timewait, we consider two
* two conditions: one in which no further processing is
* necessary (dropped || embryonic), and one in which TCP is
* not yet done, but no longer requires the socket, so the
* pcb will persist for the time being.
*
* XXXRW: Does the second case still occur?
*/
if (inp->inp_flags & INP_DROPPED ||
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
tp->t_state < TCPS_SYN_SENT) {
tcp_discardcb(tp);
in_pcbdetach(inp);
in_pcbfree(inp);
} else {
in_pcbdetach(inp);
INP_WUNLOCK(inp);
}
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
}
}
/*
* pru_detach() detaches the TCP protocol from the socket.
* If the protocol state is non-embryonic, then can't
* do this directly: have to initiate a pru_disconnect(),
* which may finish later; embryonic TCB's can just
* be discarded here.
*/
static void
tcp_usr_detach(struct socket *so)
{
struct inpcb *inp;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
INP_INFO_WLOCK(&V_tcbinfo);
INP_WLOCK(inp);
KASSERT(inp->inp_socket != NULL,
("tcp_usr_detach: inp_socket == NULL"));
tcp_detach(so, inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
}
#ifdef INET
/*
* Give the socket an address.
*/
static int
tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
struct sockaddr_in *sinp;
sinp = (struct sockaddr_in *)nam;
if (nam->sa_len != sizeof (*sinp))
return (EINVAL);
/*
* Must check for multicast addresses and disallow binding
* to them.
*/
if (sinp->sin_family == AF_INET &&
IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
return (EAFNOSUPPORT);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
error = EINVAL;
goto out;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
}
tp = intotcpcb(inp);
TCPDEBUG1();
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
error = in_pcbbind(inp, nam, td->td_ucred);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WUNLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_BIND);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#endif /* INET */
#ifdef INET6
static int
tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
struct sockaddr_in6 *sin6p;
sin6p = (struct sockaddr_in6 *)nam;
if (nam->sa_len != sizeof (*sin6p))
return (EINVAL);
/*
* Must check for multicast addresses and disallow binding
* to them.
*/
if (sin6p->sin6_family == AF_INET6 &&
IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
return (EAFNOSUPPORT);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
error = EINVAL;
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WLOCK(&V_tcbinfo);
inp->inp_vflag &= ~INP_IPV4;
inp->inp_vflag |= INP_IPV6;
#ifdef INET
if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
inp->inp_vflag |= INP_IPV4;
else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
struct sockaddr_in sin;
in6_sin6_2_sin(&sin, sin6p);
inp->inp_vflag |= INP_IPV4;
inp->inp_vflag &= ~INP_IPV6;
error = in_pcbbind(inp, (struct sockaddr *)&sin,
td->td_ucred);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WUNLOCK(&V_tcbinfo);
goto out;
}
}
#endif
error = in6_pcbbind(inp, nam, td->td_ucred);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WUNLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_BIND);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#endif /* INET6 */
#ifdef INET
/*
* Prepare to accept connections.
*/
static int
tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
error = EINVAL;
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
SOCK_LOCK(so);
error = solisten_proto_check(so);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WLOCK(&V_tcbinfo);
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
if (error == 0 && inp->inp_lport == 0)
error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WUNLOCK(&V_tcbinfo);
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
if (error == 0) {
tp->t_state = TCPS_LISTEN;
solisten_proto(so, backlog);
#ifdef TCP_OFFLOAD
tcp_offload_listen_start(tp);
#endif
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
}
SOCK_UNLOCK(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_LISTEN);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#endif /* INET */
#ifdef INET6
static int
tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
error = EINVAL;
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
SOCK_LOCK(so);
error = solisten_proto_check(so);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WLOCK(&V_tcbinfo);
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
if (error == 0 && inp->inp_lport == 0) {
inp->inp_vflag &= ~INP_IPV4;
if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
inp->inp_vflag |= INP_IPV4;
error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
}
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WUNLOCK(&V_tcbinfo);
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
if (error == 0) {
tp->t_state = TCPS_LISTEN;
solisten_proto(so, backlog);
#ifdef TCP_OFFLOAD
tcp_offload_listen_start(tp);
#endif
In the current world order, solisten() implements the state transition of a socket from a regular socket to a listening socket able to accept new connections. As part of this state transition, solisten() calls into the protocol to update protocol-layer state. There were several bugs in this implementation that could result in a race wherein a TCP SYN received in the interval between the protocol state transition and the shortly following socket layer transition would result in a panic in the TCP code, as the socket would be in the TCPS_LISTEN state, but the socket would not have the SO_ACCEPTCONN flag set. This change does the following: - Pushes the socket state transition from the socket layer solisten() to to socket "library" routines called from the protocol. This permits the socket routines to be called while holding the protocol mutexes, preventing a race exposing the incomplete socket state transition to TCP after the TCP state transition has completed. The check for a socket layer state transition is performed by solisten_proto_check(), and the actual transition is performed by solisten_proto(). - Holds the socket lock for the duration of the socket state test and set, and over the protocol layer state transition, which is now possible as the socket lock is acquired by the protocol layer, rather than vice versa. This prevents additional state related races in the socket layer. This permits the dual transition of socket layer and protocol layer state to occur while holding locks for both layers, making the two changes atomic with respect to one another. Similar changes are likely require elsewhere in the socket/protocol code. Reported by: Peter Holm <peter@holm.cc> Review and fixes from: emax, Antoine Brodin <antoine.brodin@laposte.net> Philosophical head nod: gnn
2005-02-21 21:58:17 +00:00
}
SOCK_UNLOCK(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_LISTEN);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#endif /* INET6 */
#ifdef INET
/*
* Initiate connection to peer.
* Create a template for use in transmissions on this connection.
* Enter SYN_SENT state, and mark socket as connecting.
* Start keep-alive timer, and seed output sequence space.
* Send initial segment on connection.
*/
static int
tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
struct sockaddr_in *sinp;
sinp = (struct sockaddr_in *)nam;
if (nam->sa_len != sizeof (*sinp))
return (EINVAL);
/*
* Must disallow TCP ``connections'' to multicast addresses.
*/
if (sinp->sin_family == AF_INET
&& IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
return (EAFNOSUPPORT);
if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
return (error);
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
error = EINVAL;
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
if ((error = tcp_connect(tp, nam, td)) != 0)
goto out;
#ifdef TCP_OFFLOAD
if (registered_toedevs > 0 &&
(error = tcp_offload_connect(so, nam)) == 0)
goto out;
#endif
tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
error = tcp_output(tp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_CONNECT);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#endif /* INET */
#ifdef INET6
static int
tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
struct sockaddr_in6 *sin6p;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
sin6p = (struct sockaddr_in6 *)nam;
if (nam->sa_len != sizeof (*sin6p))
return (EINVAL);
/*
* Must disallow TCP ``connections'' to multicast addresses.
*/
if (sin6p->sin6_family == AF_INET6
&& IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
return (EAFNOSUPPORT);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
error = EINVAL;
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
#ifdef INET
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
/*
* XXXRW: Some confusion: V4/V6 flags relate to binding, and
* therefore probably require the hash lock, which isn't held here.
* Is this a significant problem?
*/
if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
struct sockaddr_in sin;
if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
error = EINVAL;
goto out;
}
in6_sin6_2_sin(&sin, sin6p);
inp->inp_vflag |= INP_IPV4;
inp->inp_vflag &= ~INP_IPV6;
if ((error = prison_remote_ip4(td->td_ucred,
&sin.sin_addr)) != 0)
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
goto out;
if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
goto out;
#ifdef TCP_OFFLOAD
if (registered_toedevs > 0 &&
(error = tcp_offload_connect(so, nam)) == 0)
goto out;
#endif
error = tcp_output(tp);
goto out;
}
#endif
inp->inp_vflag &= ~INP_IPV4;
inp->inp_vflag |= INP_IPV6;
inp->inp_inc.inc_flags |= INC_ISIPV6;
if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
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
goto out;
if ((error = tcp6_connect(tp, nam, td)) != 0)
goto out;
#ifdef TCP_OFFLOAD
if (registered_toedevs > 0 &&
(error = tcp_offload_connect(so, nam)) == 0)
goto out;
#endif
tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
error = tcp_output(tp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_CONNECT);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#endif /* INET6 */
/*
* Initiate disconnect from peer.
* If connection never passed embryonic stage, just drop;
* else if don't need to let data drain, then can just drop anyways,
* else have to begin TCP shutdown process: mark socket disconnecting,
* drain unread data, state switch to reflect user close, and
* send segment (e.g. FIN) to peer. Socket will be really disconnected
* when peer sends FIN and acks ours.
*
* SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
*/
static int
tcp_usr_disconnect(struct socket *so)
{
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
int error = 0;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
INP_INFO_WLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNRESET;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
tcp_disconnect(tp);
out:
TCPDEBUG2(PRU_DISCONNECT);
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#ifdef INET
/*
* Accept a connection. Essentially all the work is done at higher levels;
* just return the address of the peer, storing through addr.
*
* The rationale for acquiring the tcbinfo lock here is somewhat complicated,
* and is described in detail in the commit log entry for r175612. Acquiring
* it delays an accept(2) racing with sonewconn(), which inserts the socket
* before the inpcb address/port fields are initialized. A better fix would
* prevent the socket from being placed in the listen queue until all fields
* are fully initialized.
*/
static int
tcp_usr_accept(struct socket *so, struct sockaddr **nam)
{
int error = 0;
struct inpcb *inp = NULL;
2001-03-12 02:57:42 +00:00
struct tcpcb *tp = NULL;
struct in_addr addr;
in_port_t port = 0;
2001-03-12 02:57:42 +00:00
TCPDEBUG0;
if (so->so_state & SS_ISDISCONNECTED)
return (ECONNABORTED);
inp = sotoinpcb(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNABORTED;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
goto out;
}
2001-03-12 02:57:42 +00:00
tp = intotcpcb(inp);
TCPDEBUG1();
/*
* We inline in_getpeeraddr and COMMON_END here, so that we can
* copy the data of interest and defer the malloc until after we
* release the lock.
*/
port = inp->inp_fport;
addr = inp->inp_faddr;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_ACCEPT);
INP_WUNLOCK(inp);
INP_INFO_RUNLOCK(&V_tcbinfo);
if (error == 0)
*nam = in_sockaddr(port, &addr);
return error;
}
#endif /* INET */
#ifdef INET6
static int
tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
{
struct inpcb *inp = NULL;
int error = 0;
2001-03-12 02:57:42 +00:00
struct tcpcb *tp = NULL;
struct in_addr addr;
struct in6_addr addr6;
in_port_t port = 0;
int v4 = 0;
2001-03-12 02:57:42 +00:00
TCPDEBUG0;
if (so->so_state & SS_ISDISCONNECTED)
return (ECONNABORTED);
inp = sotoinpcb(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_INFO_RLOCK(&V_tcbinfo);
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNABORTED;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
goto out;
}
2001-03-12 02:57:42 +00:00
tp = intotcpcb(inp);
TCPDEBUG1();
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
/*
* We inline in6_mapped_peeraddr and COMMON_END here, so that we can
* copy the data of interest and defer the malloc until after we
* release the lock.
*/
if (inp->inp_vflag & INP_IPV4) {
v4 = 1;
port = inp->inp_fport;
addr = inp->inp_faddr;
} else {
port = inp->inp_fport;
addr6 = inp->in6p_faddr;
}
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_ACCEPT);
INP_WUNLOCK(inp);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_INFO_RUNLOCK(&V_tcbinfo);
if (error == 0) {
if (v4)
*nam = in6_v4mapsin6_sockaddr(port, &addr);
else
*nam = in6_sockaddr(port, &addr6);
}
return error;
}
#endif /* INET6 */
/*
* Mark the connection as being incapable of further output.
*/
static int
tcp_usr_shutdown(struct socket *so)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
INP_INFO_WLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNRESET;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
socantsendmore(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
tcp_usrclosed(tp);
if (!(inp->inp_flags & INP_DROPPED))
error = tcp_output(tp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_SHUTDOWN);
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
/*
* After a receive, possibly send window update to peer.
*/
static int
tcp_usr_rcvd(struct socket *so, int flags)
{
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
int error = 0;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNRESET;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
#ifdef TCP_OFFLOAD
if (tp->t_flags & TF_TOE)
tcp_offload_rcvd(tp);
#endif
tcp_output(tp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_RCVD);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data. Unlike the other
* pru_*() routines, the mbuf chains are our responsibility. We
* must either enqueue them or free them. The other pru_* routines
* generally are caller-frees.
*/
static int
tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
struct sockaddr *nam, struct mbuf *control, struct thread *td)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
#ifdef INET6
int isipv6;
#endif
TCPDEBUG0;
/*
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
* We require the pcbinfo lock if we will close the socket as part of
* this call.
*/
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
if (flags & PRUS_EOF)
INP_INFO_WLOCK(&V_tcbinfo);
inp = sotoinpcb(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
if (control)
m_freem(control);
if (m)
m_freem(m);
error = ECONNRESET;
goto out;
}
#ifdef INET6
isipv6 = nam && nam->sa_family == AF_INET6;
#endif /* INET6 */
tp = intotcpcb(inp);
TCPDEBUG1();
if (control) {
/* TCP doesn't do control messages (rights, creds, etc) */
if (control->m_len) {
m_freem(control);
if (m)
m_freem(m);
error = EINVAL;
goto out;
}
m_freem(control); /* empty control, just free it */
}
if (!(flags & PRUS_OOB)) {
sbappendstream(&so->so_snd, m);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, td);
#endif /* INET6 */
#if defined(INET6) && defined(INET)
else
#endif
#ifdef INET
error = tcp_connect(tp, nam, td);
#endif
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
}
if (flags & PRUS_EOF) {
/*
* Close the send side of the connection after
* the data is sent.
*/
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
socantsendmore(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
tcp_usrclosed(tp);
}
if (!(inp->inp_flags & INP_DROPPED)) {
if (flags & PRUS_MORETOCOME)
tp->t_flags |= TF_MORETOCOME;
error = tcp_output(tp);
if (flags & PRUS_MORETOCOME)
tp->t_flags &= ~TF_MORETOCOME;
}
} else {
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
/*
* XXXRW: PRUS_EOF not implemented with PRUS_OOB?
*/
SOCKBUF_LOCK(&so->so_snd);
if (sbspace(&so->so_snd) < -512) {
SOCKBUF_UNLOCK(&so->so_snd);
m_freem(m);
error = ENOBUFS;
goto out;
}
/*
* According to RFC961 (Assigned Protocols),
* the urgent pointer points to the last octet
* of urgent data. We continue, however,
* to consider it to indicate the first octet
* of data past the urgent section.
* Otherwise, snd_up should be one lower.
*/
sbappendstream_locked(&so->so_snd, m);
SOCKBUF_UNLOCK(&so->so_snd);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, td);
#endif /* INET6 */
#if defined(INET6) && defined(INET)
else
#endif
#ifdef INET
error = tcp_connect(tp, nam, td);
#endif
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
}
tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
tp->t_flags |= TF_FORCEDATA;
error = tcp_output(tp);
tp->t_flags &= ~TF_FORCEDATA;
}
out:
TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
INP_WUNLOCK(inp);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
if (flags & PRUS_EOF)
INP_INFO_WUNLOCK(&V_tcbinfo);
return (error);
}
/*
* Abort the TCP. Drop the connection abruptly.
*/
static void
tcp_usr_abort(struct socket *so)
{
struct inpcb *inp;
struct tcpcb *tp = NULL;
TCPDEBUG0;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
INP_INFO_WLOCK(&V_tcbinfo);
INP_WLOCK(inp);
KASSERT(inp->inp_socket != NULL,
("tcp_usr_abort: inp_socket == NULL"));
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
/*
* If we still have full TCP state, and we're not dropped, drop.
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
*/
if (!(inp->inp_flags & INP_TIMEWAIT) &&
!(inp->inp_flags & INP_DROPPED)) {
tp = intotcpcb(inp);
TCPDEBUG1();
tcp_drop(tp, ECONNABORTED);
TCPDEBUG2(PRU_ABORT);
}
if (!(inp->inp_flags & INP_DROPPED)) {
SOCK_LOCK(so);
so->so_state |= SS_PROTOREF;
SOCK_UNLOCK(so);
inp->inp_flags |= INP_SOCKREF;
}
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
}
/*
* TCP socket is closed. Start friendly disconnect.
*/
static void
tcp_usr_close(struct socket *so)
{
struct inpcb *inp;
struct tcpcb *tp = NULL;
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
INP_INFO_WLOCK(&V_tcbinfo);
INP_WLOCK(inp);
KASSERT(inp->inp_socket != NULL,
("tcp_usr_close: inp_socket == NULL"));
/*
* If we still have full TCP state, and we're not dropped, initiate
* a disconnect.
*/
if (!(inp->inp_flags & INP_TIMEWAIT) &&
!(inp->inp_flags & INP_DROPPED)) {
tp = intotcpcb(inp);
TCPDEBUG1();
tcp_disconnect(tp);
TCPDEBUG2(PRU_CLOSE);
}
if (!(inp->inp_flags & INP_DROPPED)) {
SOCK_LOCK(so);
so->so_state |= SS_PROTOREF;
SOCK_UNLOCK(so);
inp->inp_flags |= INP_SOCKREF;
}
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
}
/*
* Receive out-of-band data.
*/
static int
tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
{
int error = 0;
struct inpcb *inp;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
struct tcpcb *tp = NULL;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
TCPDEBUG0;
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNRESET;
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
if ((so->so_oobmark == 0 &&
(so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
so->so_options & SO_OOBINLINE ||
tp->t_oobflags & TCPOOB_HADDATA) {
error = EINVAL;
goto out;
}
if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
error = EWOULDBLOCK;
goto out;
}
m->m_len = 1;
*mtod(m, caddr_t) = tp->t_iobc;
if ((flags & MSG_PEEK) == 0)
tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
out:
TCPDEBUG2(PRU_RCVOOB);
INP_WUNLOCK(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
return (error);
}
#ifdef INET
struct pr_usrreqs tcp_usrreqs = {
.pru_abort = tcp_usr_abort,
.pru_accept = tcp_usr_accept,
.pru_attach = tcp_usr_attach,
.pru_bind = tcp_usr_bind,
.pru_connect = tcp_usr_connect,
.pru_control = in_control,
.pru_detach = tcp_usr_detach,
.pru_disconnect = tcp_usr_disconnect,
.pru_listen = tcp_usr_listen,
.pru_peeraddr = in_getpeeraddr,
.pru_rcvd = tcp_usr_rcvd,
.pru_rcvoob = tcp_usr_rcvoob,
.pru_send = tcp_usr_send,
.pru_shutdown = tcp_usr_shutdown,
.pru_sockaddr = in_getsockaddr,
.pru_sosetlabel = in_pcbsosetlabel,
.pru_close = tcp_usr_close,
};
#endif /* INET */
1994-05-24 10:09:53 +00:00
#ifdef INET6
struct pr_usrreqs tcp6_usrreqs = {
.pru_abort = tcp_usr_abort,
.pru_accept = tcp6_usr_accept,
.pru_attach = tcp_usr_attach,
.pru_bind = tcp6_usr_bind,
.pru_connect = tcp6_usr_connect,
.pru_control = in6_control,
.pru_detach = tcp_usr_detach,
.pru_disconnect = tcp_usr_disconnect,
.pru_listen = tcp6_usr_listen,
.pru_peeraddr = in6_mapped_peeraddr,
.pru_rcvd = tcp_usr_rcvd,
.pru_rcvoob = tcp_usr_rcvoob,
.pru_send = tcp_usr_send,
.pru_shutdown = tcp_usr_shutdown,
.pru_sockaddr = in6_mapped_sockaddr,
2011-01-07 21:40:34 +00:00
.pru_sosetlabel = in_pcbsosetlabel,
.pru_close = tcp_usr_close,
};
#endif /* INET6 */
#ifdef INET
/*
* Common subroutine to open a TCP connection to remote host specified
* by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
* port number if needed. Call in_pcbconnect_setup to do the routing and
* to choose a local host address (interface). If there is an existing
* incarnation of the same connection in TIME-WAIT state and if the remote
* host was sending CC options and if the connection duration was < MSL, then
* truncate the previous TIME-WAIT state and proceed.
* Initialize connection parameters and enter SYN-SENT state.
*/
static int
tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
{
struct inpcb *inp = tp->t_inpcb, *oinp;
struct socket *so = inp->inp_socket;
struct in_addr laddr;
u_short lport;
Improved connection establishment performance by doing local port lookups via a hashed port list. In the new scheme, in_pcblookup() goes away and is replaced by a new routine, in_pcblookup_local() for doing the local port check. Note that this implementation is space inefficient in that the PCB struct is now too large to fit into 128 bytes. I might deal with this in the future by using the new zone allocator, but I wanted these changes to be extensively tested in their current form first. Also: 1) Fixed off-by-one errors in the port lookup loops in in_pcbbind(). 2) Got rid of some unneeded rehashing. Adding a new routine, in_pcbinshash() to do the initialial hash insertion. 3) Renamed in_pcblookuphash() to in_pcblookup_hash() for easier readability. 4) Added a new routine, in_pcbremlists() to remove the PCB from the various hash lists. 5) Added/deleted comments where appropriate. 6) Removed unnecessary splnet() locking. In general, the PCB functions should be called at splnet()...there are unfortunately a few exceptions, however. 7) Reorganized a few structs for better cache line behavior. 8) Killed my TCP_ACK_HACK kludge. It may come back in a different form in the future, however. These changes have been tested on wcarchive for more than a month. In tests done here, connection establishment overhead is reduced by more than 50 times, thus getting rid of one of the major networking scalability problems. Still to do: make tcp_fastimo/tcp_slowtimo scale well for systems with a large number of connections. tcp_fastimo is easy; tcp_slowtimo is difficult. WARNING: Anything that knows about inpcb and tcpcb structs will have to be recompiled; at the very least, this includes netstat(1).
1998-01-27 09:15:13 +00:00
int error;
INP_WLOCK_ASSERT(inp);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
if (inp->inp_lport == 0) {
error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
if (error)
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
goto out;
}
1995-05-30 08:16:23 +00:00
/*
* Cannot simply call in_pcbconnect, because there might be an
* earlier incarnation of this same connection still in
* TIME_WAIT state, creating an ADDRINUSE error.
*/
laddr = inp->inp_laddr;
lport = inp->inp_lport;
error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
&inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
if (error && oinp == NULL)
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
goto out;
if (oinp) {
error = EADDRINUSE;
goto out;
}
inp->inp_laddr = laddr;
in_pcbrehash(inp);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WUNLOCK(&V_tcbinfo);
/*
* Compute window scaling to request:
* Scale to fit into sweet spot. See tcp_syncache.c.
* XXX: This should move to tcp_output().
*/
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
(TCP_MAXWIN << tp->request_r_scale) < sb_max)
tp->request_r_scale++;
soisconnecting(so);
TCPSTAT_INC(tcps_connattempt);
tp->t_state = TCPS_SYN_SENT;
tp->iss = tcp_new_isn(tp);
tcp_sendseqinit(tp);
return 0;
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
out:
INP_HASH_WUNLOCK(&V_tcbinfo);
return (error);
}
#endif /* INET */
#ifdef INET6
static int
tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
{
struct inpcb *inp = tp->t_inpcb, *oinp;
struct socket *so = inp->inp_socket;
struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
struct in6_addr addr6;
int error;
INP_WLOCK_ASSERT(inp);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WLOCK(&V_tcbinfo);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
if (inp->inp_lport == 0) {
error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
if (error)
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
goto out;
}
/*
* Cannot simply call in_pcbconnect, because there might be an
* earlier incarnation of this same connection still in
* TIME_WAIT state, creating an ADDRINUSE error.
* in6_pcbladdr() also handles scope zone IDs.
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
*
* XXXRW: We wouldn't need to expose in6_pcblookup_hash_locked()
* outside of in6_pcb.c if there were an in6_pcbconnect_setup().
*/
error = in6_pcbladdr(inp, nam, &addr6);
if (error)
goto out;
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
oinp = in6_pcblookup_hash_locked(inp->inp_pcbinfo,
&sin6->sin6_addr, sin6->sin6_port,
IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
? &addr6
: &inp->in6p_laddr,
inp->inp_lport, 0, NULL);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
if (oinp) {
error = EADDRINUSE;
goto out;
}
if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
inp->in6p_laddr = addr6;
inp->in6p_faddr = sin6->sin6_addr;
inp->inp_fport = sin6->sin6_port;
/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
inp->inp_flow |=
(htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
in_pcbrehash(inp);
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
INP_HASH_WUNLOCK(&V_tcbinfo);
/* Compute window scaling to request. */
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
(TCP_MAXWIN << tp->request_r_scale) < sb_max)
tp->request_r_scale++;
soisconnecting(so);
TCPSTAT_INC(tcps_connattempt);
tp->t_state = TCPS_SYN_SENT;
tp->iss = tcp_new_isn(tp);
tcp_sendseqinit(tp);
return 0;
Decompose the current single inpcbinfo lock into two locks: - The existing ipi_lock continues to protect the global inpcb list and inpcb counter. This lock is now relegated to a small number of allocation and free operations, and occasional operations that walk all connections (including, awkwardly, certain UDP multicast receive operations -- something to revisit). - A new ipi_hash_lock protects the two inpcbinfo hash tables for looking up connections and bound sockets, manipulated using new INP_HASH_*() macros. This lock, combined with inpcb locks, protects the 4-tuple address space. Unlike the current ipi_lock, ipi_hash_lock follows the individual inpcb connection locks, so may be acquired while manipulating a connection on which a lock is already held, avoiding the need to acquire the inpcbinfo lock preemptively when a binding change might later be required. As a result, however, lookup operations necessarily go through a reference acquire while holding the lookup lock, later acquiring an inpcb lock -- if required. A new function in_pcblookup() looks up connections, and accepts flags indicating how to return the inpcb. Due to lock order changes, callers no longer need acquire locks before performing a lookup: the lookup routine will acquire the ipi_hash_lock as needed. In the future, it will also be able to use alternative lookup and locking strategies transparently to callers, such as pcbgroup lookup. New lookup flags are, supplementing the existing INPLOOKUP_WILDCARD flag: INPLOOKUP_RLOCKPCB - Acquire a read lock on the returned inpcb INPLOOKUP_WLOCKPCB - Acquire a write lock on the returned inpcb Callers must pass exactly one of these flags (for the time being). Some notes: - All protocols are updated to work within the new regime; especially, TCP, UDPv4, and UDPv6. pcbinfo ipi_lock acquisitions are largely eliminated, and global hash lock hold times are dramatically reduced compared to previous locking. - The TCP syncache still relies on the pcbinfo lock, something that we may want to revisit. - Support for reverting to the FreeBSD 7.x locking strategy in TCP input is no longer available -- hash lookup locks are now held only very briefly during inpcb lookup, rather than for potentially extended periods. However, the pcbinfo ipi_lock will still be acquired if a connection state might change such that a connection is added or removed. - Raw IP sockets continue to use the pcbinfo ipi_lock for protection, due to maintaining their own hash tables. - The interface in6_pcblookup_hash_locked() is maintained, which allows callers to acquire hash locks and perform one or more lookups atomically with 4-tuple allocation: this is required only for TCPv6, as there is no in6_pcbconnect_setup(), which there should be. - UDPv6 locking remains significantly more conservative than UDPv4 locking, which relates to source address selection. This needs attention, as it likely significantly reduces parallelism in this code for multithreaded socket use (such as in BIND). - In the UDPv4 and UDPv6 multicast cases, we need to revisit locking somewhat, as they relied on ipi_lock to stablise 4-tuple matches, which is no longer sufficient. A second check once the inpcb lock is held should do the trick, keeping the general case from requiring the inpcb lock for every inpcb visited. - This work reminds us that we need to revisit locking of the v4/v6 flags, which may be accessed lock-free both before and after this change. - Right now, a single lock name is used for the pcbhash lock -- this is undesirable, and probably another argument is required to take care of this (or a char array name field in the pcbinfo?). This is not an MFC candidate for 8.x due to its impact on lookup and locking semantics. It's possible some of these issues could be worked around with compatibility wrappers, if necessary. Reviewed by: bz Sponsored by: Juniper Networks, Inc.
2011-05-30 09:43:55 +00:00
out:
INP_HASH_WUNLOCK(&V_tcbinfo);
return error;
}
#endif /* INET6 */
/*
* Export TCP internal state information via a struct tcp_info, based on the
* Linux 2.6 API. Not ABI compatible as our constants are mapped differently
* (TCP state machine, etc). We export all information using FreeBSD-native
* constants -- for example, the numeric values for tcpi_state will differ
* from Linux.
*/
static void
tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
{
INP_WLOCK_ASSERT(tp->t_inpcb);
bzero(ti, sizeof(*ti));
ti->tcpi_state = tp->t_state;
if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
if (tp->t_flags & TF_SACK_PERMIT)
ti->tcpi_options |= TCPI_OPT_SACK;
if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
ti->tcpi_options |= TCPI_OPT_WSCALE;
ti->tcpi_snd_wscale = tp->snd_scale;
ti->tcpi_rcv_wscale = tp->rcv_scale;
}
ti->tcpi_rto = tp->t_rxtcur * tick;
ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
ti->tcpi_snd_cwnd = tp->snd_cwnd;
/*
* FreeBSD-specific extension fields for tcp_info.
*/
ti->tcpi_rcv_space = tp->rcv_wnd;
ti->tcpi_rcv_nxt = tp->rcv_nxt;
ti->tcpi_snd_wnd = tp->snd_wnd;
ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
ti->tcpi_snd_nxt = tp->snd_nxt;
ti->tcpi_snd_mss = tp->t_maxseg;
ti->tcpi_rcv_mss = tp->t_maxseg;
if (tp->t_flags & TF_TOE)
ti->tcpi_options |= TCPI_OPT_TOE;
ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
ti->tcpi_snd_zerowin = tp->t_sndzerowin;
}
/*
* tcp_ctloutput() must drop the inpcb lock before performing copyin on
* socket option arguments. When it re-acquires the lock after the copy, it
* has to revalidate that the connection is still valid for the socket
* option.
*/
#define INP_WLOCK_RECHECK(inp) do { \
INP_WLOCK(inp); \
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
INP_WUNLOCK(inp); \
return (ECONNRESET); \
} \
tp = intotcpcb(inp); \
} while(0)
1994-05-24 10:09:53 +00:00
int
tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1994-05-24 10:09:53 +00:00
{
int error, opt, optval;
u_int ui;
struct inpcb *inp;
struct tcpcb *tp;
struct tcp_info ti;
This commit marks the first formal contribution of the "Five New TCP Congestion Control Algorithms for FreeBSD" FreeBSD Foundation funded project. More details about the project are available at: http://caia.swin.edu.au/freebsd/5cc/ - Add a KPI and supporting infrastructure to allow modular congestion control algorithms to be used in the net stack. Algorithms can maintain per-connection state if required, and connections maintain their own algorithm pointer, which allows different connections to concurrently use different algorithms. The TCP_CONGESTION socket option can be used with getsockopt()/setsockopt() to programmatically query or change the congestion control algorithm respectively from within an application at runtime. - Integrate the framework with the TCP stack in as least intrusive a manner as possible. Care was also taken to develop the framework in a way that should allow integration with other congestion aware transport protocols (e.g. SCTP) in the future. The hope is that we will one day be able to share a single set of congestion control algorithm modules between all congestion aware transport protocols. - Introduce a new congestion recovery (TF_CONGRECOVERY) state into the TCP stack and use it to decouple the meaning of recovery from a congestion event and recovery from packet loss (TF_FASTRECOVERY) a la RFC2581. ECN and delay based congestion control protocols don't generally need to recover from packet loss and need a different way to note a congestion recovery episode within the stack. - Remove the net.inet.tcp.newreno sysctl, which simplifies some portions of code and ensures the stack always uses the appropriate mechanisms for recovering from packet loss during a congestion recovery episode. - Extract the NewReno congestion control algorithm from the TCP stack and massage it into module form. NewReno is always built into the kernel and will remain the default algorithm for the forseeable future. Implementations of additional different algorithms will become available in the near future. - Bump __FreeBSD_version to 900025 and note in UPDATING that rebuilding code that relies on the size of "struct tcpcb" is required. Many thanks go to the Cisco University Research Program Fund at Community Foundation Silicon Valley and the FreeBSD Foundation. Their support of our work at the Centre for Advanced Internet Architectures, Swinburne University of Technology is greatly appreciated. In collaboration with: David Hayes <dahayes at swin edu au> and Grenville Armitage <garmitage at swin edu au> Sponsored by: Cisco URP, FreeBSD Foundation Reviewed by: rpaulo Tested by: David Hayes (and many others over the years) MFC after: 3 months
2010-11-12 06:41:55 +00:00
char buf[TCP_CA_NAME_MAX];
struct cc_algo *algo;
1994-05-24 10:09:53 +00:00
error = 0;
1994-05-24 10:09:53 +00:00
inp = sotoinpcb(so);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
INP_WLOCK(inp);
if (sopt->sopt_level != IPPROTO_TCP) {
#ifdef INET6
if (inp->inp_vflag & INP_IPV6PROTO) {
INP_WUNLOCK(inp);
error = ip6_ctloutput(so, sopt);
}
#endif /* INET6 */
#if defined(INET6) && defined(INET)
else
#endif
#ifdef INET
{
INP_WUNLOCK(inp);
error = ip_ctloutput(so, sopt);
}
#endif
1994-05-24 10:09:53 +00:00
return (error);
}
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
INP_WUNLOCK(inp);
return (ECONNRESET);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
}
1994-05-24 10:09:53 +00:00
switch (sopt->sopt_dir) {
case SOPT_SET:
switch (sopt->sopt_name) {
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#ifdef TCP_SIGNATURE
case TCP_MD5SIG:
INP_WUNLOCK(inp);
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
error = sooptcopyin(sopt, &optval, sizeof optval,
sizeof optval);
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
if (error)
return (error);
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
INP_WLOCK_RECHECK(inp);
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
if (optval > 0)
tp->t_flags |= TF_SIGNATURE;
else
tp->t_flags &= ~TF_SIGNATURE;
goto unlock_and_done;
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#endif /* TCP_SIGNATURE */
1994-05-24 10:09:53 +00:00
case TCP_NODELAY:
case TCP_NOOPT:
INP_WUNLOCK(inp);
error = sooptcopyin(sopt, &optval, sizeof optval,
sizeof optval);
if (error)
return (error);
INP_WLOCK_RECHECK(inp);
switch (sopt->sopt_name) {
case TCP_NODELAY:
opt = TF_NODELAY;
break;
case TCP_NOOPT:
opt = TF_NOOPT;
break;
default:
opt = 0; /* dead code to fool gcc */
break;
}
if (optval)
tp->t_flags |= opt;
1994-05-24 10:09:53 +00:00
else
tp->t_flags &= ~opt;
unlock_and_done:
#ifdef TCP_OFFLOAD
if (tp->t_flags & TF_TOE) {
tcp_offload_ctloutput(tp, sopt->sopt_dir,
sopt->sopt_name);
}
#endif
INP_WUNLOCK(inp);
1994-05-24 10:09:53 +00:00
break;
case TCP_NOPUSH:
INP_WUNLOCK(inp);
error = sooptcopyin(sopt, &optval, sizeof optval,
sizeof optval);
if (error)
return (error);
INP_WLOCK_RECHECK(inp);
if (optval)
tp->t_flags |= TF_NOPUSH;
else if (tp->t_flags & TF_NOPUSH) {
tp->t_flags &= ~TF_NOPUSH;
if (TCPS_HAVEESTABLISHED(tp->t_state))
error = tcp_output(tp);
}
goto unlock_and_done;
1994-05-24 10:09:53 +00:00
case TCP_MAXSEG:
INP_WUNLOCK(inp);
error = sooptcopyin(sopt, &optval, sizeof optval,
sizeof optval);
if (error)
return (error);
1994-05-24 10:09:53 +00:00
INP_WLOCK_RECHECK(inp);
Limiters and sanity checks for TCP MSS (maximum segement size) resource exhaustion attacks. For network link optimization TCP can adjust its MSS and thus packet size according to the observed path MTU. This is done dynamically based on feedback from the remote host and network components along the packet path. This information can be abused to pretend an extremely low path MTU. The resource exhaustion works in two ways: o during tcp connection setup the advertized local MSS is exchanged between the endpoints. The remote endpoint can set this arbitrarily low (except for a minimum MTU of 64 octets enforced in the BSD code). When the local host is sending data it is forced to send many small IP packets instead of a large one. For example instead of the normal TCP payload size of 1448 it forces TCP payload size of 12 (MTU 64) and thus we have a 120 times increase in workload and packets. On fast links this quickly saturates the local CPU and may also hit pps processing limites of network components along the path. This type of attack is particularly effective for servers where the attacker can download large files (WWW and FTP). We mitigate it by enforcing a minimum MTU settable by sysctl net.inet.tcp.minmss defaulting to 256 octets. o the local host is reveiving data on a TCP connection from the remote host. The local host has no control over the packet size the remote host is sending. The remote host may chose to do what is described in the first attack and send the data in packets with an TCP payload of at least one byte. For each packet the tcp_input() function will be entered, the packet is processed and a sowakeup() is signalled to the connected process. For example an attack with 2 Mbit/s gives 4716 packets per second and the same amount of sowakeup()s to the process (and context switches). This type of attack is particularly effective for servers where the attacker can upload large amounts of data. Normally this is the case with WWW server where large POSTs can be made. We mitigate this by calculating the average MSS payload per second. If it goes below 'net.inet.tcp.minmss' and the pps rate is above 'net.inet.tcp.minmssoverload' defaulting to 1000 this particular TCP connection is resetted and dropped. MITRE CVE: CAN-2004-0002 Reviewed by: sam (mentor) MFC after: 1 day
2004-01-08 17:40:07 +00:00
if (optval > 0 && optval <= tp->t_maxseg &&
optval + 40 >= V_tcp_minmss)
tp->t_maxseg = optval;
else
error = EINVAL;
goto unlock_and_done;
case TCP_INFO:
INP_WUNLOCK(inp);
error = EINVAL;
break;
This commit marks the first formal contribution of the "Five New TCP Congestion Control Algorithms for FreeBSD" FreeBSD Foundation funded project. More details about the project are available at: http://caia.swin.edu.au/freebsd/5cc/ - Add a KPI and supporting infrastructure to allow modular congestion control algorithms to be used in the net stack. Algorithms can maintain per-connection state if required, and connections maintain their own algorithm pointer, which allows different connections to concurrently use different algorithms. The TCP_CONGESTION socket option can be used with getsockopt()/setsockopt() to programmatically query or change the congestion control algorithm respectively from within an application at runtime. - Integrate the framework with the TCP stack in as least intrusive a manner as possible. Care was also taken to develop the framework in a way that should allow integration with other congestion aware transport protocols (e.g. SCTP) in the future. The hope is that we will one day be able to share a single set of congestion control algorithm modules between all congestion aware transport protocols. - Introduce a new congestion recovery (TF_CONGRECOVERY) state into the TCP stack and use it to decouple the meaning of recovery from a congestion event and recovery from packet loss (TF_FASTRECOVERY) a la RFC2581. ECN and delay based congestion control protocols don't generally need to recover from packet loss and need a different way to note a congestion recovery episode within the stack. - Remove the net.inet.tcp.newreno sysctl, which simplifies some portions of code and ensures the stack always uses the appropriate mechanisms for recovering from packet loss during a congestion recovery episode. - Extract the NewReno congestion control algorithm from the TCP stack and massage it into module form. NewReno is always built into the kernel and will remain the default algorithm for the forseeable future. Implementations of additional different algorithms will become available in the near future. - Bump __FreeBSD_version to 900025 and note in UPDATING that rebuilding code that relies on the size of "struct tcpcb" is required. Many thanks go to the Cisco University Research Program Fund at Community Foundation Silicon Valley and the FreeBSD Foundation. Their support of our work at the Centre for Advanced Internet Architectures, Swinburne University of Technology is greatly appreciated. In collaboration with: David Hayes <dahayes at swin edu au> and Grenville Armitage <garmitage at swin edu au> Sponsored by: Cisco URP, FreeBSD Foundation Reviewed by: rpaulo Tested by: David Hayes (and many others over the years) MFC after: 3 months
2010-11-12 06:41:55 +00:00
case TCP_CONGESTION:
INP_WUNLOCK(inp);
bzero(buf, sizeof(buf));
error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
if (error)
break;
INP_WLOCK_RECHECK(inp);
/*
* Return EINVAL if we can't find the requested cc algo.
*/
error = EINVAL;
CC_LIST_RLOCK();
STAILQ_FOREACH(algo, &cc_list, entries) {
if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
== 0) {
/* We've found the requested algo. */
error = 0;
/*
* We hold a write lock over the tcb
* so it's safe to do these things
* without ordering concerns.
*/
if (CC_ALGO(tp)->cb_destroy != NULL)
CC_ALGO(tp)->cb_destroy(tp->ccv);
CC_ALGO(tp) = algo;
/*
* If something goes pear shaped
* initialising the new algo,
* fall back to newreno (which
* does not require initialisation).
*/
if (algo->cb_init != NULL)
if (algo->cb_init(tp->ccv) > 0) {
CC_ALGO(tp) = &newreno_cc_algo;
/*
* The only reason init
* should fail is
* because of malloc.
*/
error = ENOMEM;
}
break; /* Break the STAILQ_FOREACH. */
}
}
CC_LIST_RUNLOCK();
goto unlock_and_done;
This commit marks the first formal contribution of the "Five New TCP Congestion Control Algorithms for FreeBSD" FreeBSD Foundation funded project. More details about the project are available at: http://caia.swin.edu.au/freebsd/5cc/ - Add a KPI and supporting infrastructure to allow modular congestion control algorithms to be used in the net stack. Algorithms can maintain per-connection state if required, and connections maintain their own algorithm pointer, which allows different connections to concurrently use different algorithms. The TCP_CONGESTION socket option can be used with getsockopt()/setsockopt() to programmatically query or change the congestion control algorithm respectively from within an application at runtime. - Integrate the framework with the TCP stack in as least intrusive a manner as possible. Care was also taken to develop the framework in a way that should allow integration with other congestion aware transport protocols (e.g. SCTP) in the future. The hope is that we will one day be able to share a single set of congestion control algorithm modules between all congestion aware transport protocols. - Introduce a new congestion recovery (TF_CONGRECOVERY) state into the TCP stack and use it to decouple the meaning of recovery from a congestion event and recovery from packet loss (TF_FASTRECOVERY) a la RFC2581. ECN and delay based congestion control protocols don't generally need to recover from packet loss and need a different way to note a congestion recovery episode within the stack. - Remove the net.inet.tcp.newreno sysctl, which simplifies some portions of code and ensures the stack always uses the appropriate mechanisms for recovering from packet loss during a congestion recovery episode. - Extract the NewReno congestion control algorithm from the TCP stack and massage it into module form. NewReno is always built into the kernel and will remain the default algorithm for the forseeable future. Implementations of additional different algorithms will become available in the near future. - Bump __FreeBSD_version to 900025 and note in UPDATING that rebuilding code that relies on the size of "struct tcpcb" is required. Many thanks go to the Cisco University Research Program Fund at Community Foundation Silicon Valley and the FreeBSD Foundation. Their support of our work at the Centre for Advanced Internet Architectures, Swinburne University of Technology is greatly appreciated. In collaboration with: David Hayes <dahayes at swin edu au> and Grenville Armitage <garmitage at swin edu au> Sponsored by: Cisco URP, FreeBSD Foundation Reviewed by: rpaulo Tested by: David Hayes (and many others over the years) MFC after: 3 months
2010-11-12 06:41:55 +00:00
case TCP_KEEPIDLE:
case TCP_KEEPINTVL:
case TCP_KEEPCNT:
case TCP_KEEPINIT:
INP_WUNLOCK(inp);
error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
if (error)
return (error);
if (ui > (UINT_MAX / hz)) {
error = EINVAL;
break;
}
ui *= hz;
INP_WLOCK_RECHECK(inp);
switch (sopt->sopt_name) {
case TCP_KEEPIDLE:
tp->t_keepidle = ui;
/*
* XXX: better check current remaining
* timeout and "merge" it with new value.
*/
if ((tp->t_state > TCPS_LISTEN) &&
(tp->t_state <= TCPS_CLOSING))
tcp_timer_activate(tp, TT_KEEP,
TP_KEEPIDLE(tp));
break;
case TCP_KEEPINTVL:
tp->t_keepintvl = ui;
if ((tp->t_state == TCPS_FIN_WAIT_2) &&
(TP_MAXIDLE(tp) > 0))
tcp_timer_activate(tp, TT_2MSL,
TP_MAXIDLE(tp));
break;
case TCP_KEEPCNT:
tp->t_keepcnt = ui;
if ((tp->t_state == TCPS_FIN_WAIT_2) &&
(TP_MAXIDLE(tp) > 0))
tcp_timer_activate(tp, TT_2MSL,
TP_MAXIDLE(tp));
break;
case TCP_KEEPINIT:
tp->t_keepinit = ui;
if (tp->t_state == TCPS_SYN_RECEIVED ||
tp->t_state == TCPS_SYN_SENT)
tcp_timer_activate(tp, TT_KEEP,
TP_KEEPINIT(tp));
break;
}
goto unlock_and_done;
1994-05-24 10:09:53 +00:00
default:
INP_WUNLOCK(inp);
1994-05-24 10:09:53 +00:00
error = ENOPROTOOPT;
break;
}
break;
case SOPT_GET:
tp = intotcpcb(inp);
switch (sopt->sopt_name) {
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
#ifdef TCP_SIGNATURE
case TCP_MD5SIG:
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);
Initial import of RFC 2385 (TCP-MD5) digest support. This is the first of two commits; bringing in the kernel support first. This can be enabled by compiling a kernel with options TCP_SIGNATURE and FAST_IPSEC. For the uninitiated, this is a TCP option which provides for a means of authenticating TCP sessions which came into being before IPSEC. It is still relevant today, however, as it is used by many commercial router vendors, particularly with BGP, and as such has become a requirement for interconnect at many major Internet points of presence. Several parts of the TCP and IP headers, including the segment payload, are digested with MD5, including a shared secret. The PF_KEY interface is used to manage the secrets using security associations in the SADB. There is a limitation here in that as there is no way to map a TCP flow per-port back to an SPI without polluting tcpcb or using the SPD; the code to do the latter is unstable at this time. Therefore this code only supports per-host keying granularity. Whilst FAST_IPSEC is mutually exclusive with KAME IPSEC (and thus IPv6), TCP_SIGNATURE applies only to IPv4. For the vast majority of prospective users of this feature, this will not pose any problem. This implementation is output-only; that is, the option is honoured when responding to a host initiating a TCP session, but no effort is made [yet] to authenticate inbound traffic. This is, however, sufficient to interwork with Cisco equipment. Tested with a Cisco 2501 running IOS 12.0(27), and Quagga 0.96.4 with local patches. Patches for tcpdump to validate TCP-MD5 sessions are also available from me upon request. Sponsored by: sentex.net
2004-02-11 04:26:04 +00:00
break;
2004-02-13 18:21:45 +00:00
#endif
1994-05-24 10:09:53 +00:00
case TCP_NODELAY:
optval = tp->t_flags & TF_NODELAY;
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);
1994-05-24 10:09:53 +00:00
break;
case TCP_MAXSEG:
optval = tp->t_maxseg;
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);
1994-05-24 10:09:53 +00:00
break;
case TCP_NOOPT:
optval = tp->t_flags & TF_NOOPT;
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);
break;
case TCP_NOPUSH:
optval = tp->t_flags & TF_NOPUSH;
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &optval, sizeof optval);
break;
case TCP_INFO:
tcp_fill_info(tp, &ti);
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, &ti, sizeof ti);
break;
This commit marks the first formal contribution of the "Five New TCP Congestion Control Algorithms for FreeBSD" FreeBSD Foundation funded project. More details about the project are available at: http://caia.swin.edu.au/freebsd/5cc/ - Add a KPI and supporting infrastructure to allow modular congestion control algorithms to be used in the net stack. Algorithms can maintain per-connection state if required, and connections maintain their own algorithm pointer, which allows different connections to concurrently use different algorithms. The TCP_CONGESTION socket option can be used with getsockopt()/setsockopt() to programmatically query or change the congestion control algorithm respectively from within an application at runtime. - Integrate the framework with the TCP stack in as least intrusive a manner as possible. Care was also taken to develop the framework in a way that should allow integration with other congestion aware transport protocols (e.g. SCTP) in the future. The hope is that we will one day be able to share a single set of congestion control algorithm modules between all congestion aware transport protocols. - Introduce a new congestion recovery (TF_CONGRECOVERY) state into the TCP stack and use it to decouple the meaning of recovery from a congestion event and recovery from packet loss (TF_FASTRECOVERY) a la RFC2581. ECN and delay based congestion control protocols don't generally need to recover from packet loss and need a different way to note a congestion recovery episode within the stack. - Remove the net.inet.tcp.newreno sysctl, which simplifies some portions of code and ensures the stack always uses the appropriate mechanisms for recovering from packet loss during a congestion recovery episode. - Extract the NewReno congestion control algorithm from the TCP stack and massage it into module form. NewReno is always built into the kernel and will remain the default algorithm for the forseeable future. Implementations of additional different algorithms will become available in the near future. - Bump __FreeBSD_version to 900025 and note in UPDATING that rebuilding code that relies on the size of "struct tcpcb" is required. Many thanks go to the Cisco University Research Program Fund at Community Foundation Silicon Valley and the FreeBSD Foundation. Their support of our work at the Centre for Advanced Internet Architectures, Swinburne University of Technology is greatly appreciated. In collaboration with: David Hayes <dahayes at swin edu au> and Grenville Armitage <garmitage at swin edu au> Sponsored by: Cisco URP, FreeBSD Foundation Reviewed by: rpaulo Tested by: David Hayes (and many others over the years) MFC after: 3 months
2010-11-12 06:41:55 +00:00
case TCP_CONGESTION:
bzero(buf, sizeof(buf));
strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
INP_WUNLOCK(inp);
error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
break;
1994-05-24 10:09:53 +00:00
default:
INP_WUNLOCK(inp);
1994-05-24 10:09:53 +00:00
error = ENOPROTOOPT;
break;
}
break;
}
return (error);
}
#undef INP_WLOCK_RECHECK
1994-05-24 10:09:53 +00:00
/*
* Attach TCP protocol to socket, allocating
* internet protocol control block, tcp control block,
* bufer space, and entering LISTEN state if to accept connections.
*/
static int
tcp_attach(struct socket *so)
1994-05-24 10:09:53 +00:00
{
struct tcpcb *tp;
1994-05-24 10:09:53 +00:00
struct inpcb *inp;
int error;
if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
1994-05-24 10:09:53 +00:00
if (error)
return (error);
}
so->so_rcv.sb_flags |= SB_AUTOSIZE;
so->so_snd.sb_flags |= SB_AUTOSIZE;
INP_INFO_WLOCK(&V_tcbinfo);
error = in_pcballoc(so, &V_tcbinfo);
if (error) {
INP_INFO_WUNLOCK(&V_tcbinfo);
1994-05-24 10:09:53 +00:00
return (error);
}
1994-05-24 10:09:53 +00:00
inp = sotoinpcb(so);
#ifdef INET6
if (inp->inp_vflag & INP_IPV6PROTO) {
inp->inp_vflag |= INP_IPV6;
inp->in6p_hops = -1; /* use kernel default */
}
else
#endif
inp->inp_vflag |= INP_IPV4;
1994-05-24 10:09:53 +00:00
tp = tcp_newtcpcb(inp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
if (tp == NULL) {
in_pcbdetach(inp);
in_pcbfree(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
1994-05-24 10:09:53 +00:00
return (ENOBUFS);
}
tp->t_state = TCPS_CLOSED;
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
1994-05-24 10:09:53 +00:00
return (0);
}
/*
* Initiate (or continue) disconnect.
* If embryonic state, just send reset (once).
* If in ``let data drain'' option and linger null, just drop.
* Otherwise (hard), mark socket disconnecting and drop
* current input data; switch states based on user close, and
* send segment to peer (with FIN).
*/
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
static void
tcp_disconnect(struct tcpcb *tp)
1994-05-24 10:09:53 +00:00
{
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
INP_WLOCK_ASSERT(inp);
1994-05-24 10:09:53 +00:00
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
/*
* Neither tcp_close() nor tcp_drop() should return NULL, as the
* socket is still open.
*/
if (tp->t_state < TCPS_ESTABLISHED) {
1994-05-24 10:09:53 +00:00
tp = tcp_close(tp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
KASSERT(tp != NULL,
("tcp_disconnect: tcp_close() returned NULL"));
} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
tp = tcp_drop(tp, 0);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
KASSERT(tp != NULL,
("tcp_disconnect: tcp_drop() returned NULL"));
} else {
soisdisconnecting(so);
sbflush(&so->so_rcv);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
tcp_usrclosed(tp);
if (!(inp->inp_flags & INP_DROPPED))
tcp_output(tp);
1994-05-24 10:09:53 +00:00
}
}
/*
* User issued close, and wish to trail through shutdown states:
* if never received SYN, just forget it. If got a SYN from peer,
* but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
* If already got a FIN from peer, then almost done; go to LAST_ACK
* state. In all other cases, have already sent FIN to peer (e.g.
* after PRU_SHUTDOWN), and just have to play tedious game waiting
* for peer to send FIN or not respond to keep-alives, etc.
* We can let the user exit from the close as soon as the FIN is acked.
*/
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
static void
tcp_usrclosed(struct tcpcb *tp)
1994-05-24 10:09:53 +00:00
{
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
INP_WLOCK_ASSERT(tp->t_inpcb);
1994-05-24 10:09:53 +00:00
switch (tp->t_state) {
case TCPS_LISTEN:
#ifdef TCP_OFFLOAD
tcp_offload_listen_stop(tp);
#endif
/* FALLTHROUGH */
case TCPS_CLOSED:
1994-05-24 10:09:53 +00:00
tp->t_state = TCPS_CLOSED;
tp = tcp_close(tp);
Update TCP for infrastructural changes to the socket/pcb refcount model, pru_abort(), pru_detach(), and in_pcbdetach(): - Universally support and enforce the invariant that so_pcb is never NULL, converting dozens of unnecessary NULL checks into assertions, and eliminating dozens of unnecessary error handling cases in protocol code. - In some cases, eliminate unnecessary pcbinfo locking, as it is no longer required to ensure so_pcb != NULL. For example, the receive code no longer requires the pcbinfo lock, and the send code only requires it if building a new connection on an otherwise unconnected socket triggered via sendto() with an address. This should significnatly reduce tcbinfo lock contention in the receive and send cases. - In order to support the invariant that so_pcb != NULL, it is now necessary for the TCP code to not discard the tcpcb any time a connection is dropped, but instead leave the tcpcb until the socket is shutdown. This case is handled by setting INP_DROPPED, to substitute for using a NULL so_pcb to indicate that the connection has been dropped. This requires the inpcb lock, but not the pcbinfo lock. - Unlike all other protocols in the tree, TCP may need to retain access to the socket after the file descriptor has been closed. Set SS_PROTOREF in tcp_detach() in order to prevent the socket from being freed, and add a flag, INP_SOCKREF, so that the TCP code knows whether or not it needs to free the socket when the connection finally does close. The typical case where this occurs is if close() is called on a TCP socket before all sent data in the send socket buffer has been transmitted or acknowledged. If INP_SOCKREF is found when the connection is dropped, we release the inpcb, tcpcb, and socket instead of flagging INP_DROPPED. - Abort and detach protocol switch methods no longer return failures, nor attempt to free sockets, as the socket layer does this. - Annotate the existence of a long-standing race in the TCP timer code, in which timers are stopped but not drained when the socket is freed, as waiting for drain may lead to deadlocks, or have to occur in a context where waiting is not permitted. This race has been handled by testing to see if the tcpcb pointer in the inpcb is NULL (and vice versa), which is not normally permitted, but may be true of a inpcb and tcpcb have been freed. Add a counter to test how often this race has actually occurred, and a large comment for each instance where we compare potentially freed memory with NULL. This will have to be fixed in the near future, but requires is to further address how to handle the timer shutdown shutdown issue. - Several TCP calls no longer potentially free the passed inpcb/tcpcb, so no longer need to return a pointer to indicate whether the argument passed in is still valid. - Un-macroize debugging and locking setup for various protocol switch methods for TCP, as it lead to more obscurity, and as locking becomes more customized to the methods, offers less benefit. - Assert copyright on tcp_usrreq.c due to significant modifications that have been made as part of this work. These changes significantly modify the memory management and connection logic of our TCP implementation, and are (as such) High Risk Changes, and likely to contain serious bugs. Please report problems to the current@ mailing list ASAP, ideally with simple test cases, and optionally, packet traces. MFC after: 3 months
2006-04-01 16:36:36 +00:00
/*
* tcp_close() should never return NULL here as the socket is
* still open.
*/
KASSERT(tp != NULL,
("tcp_usrclosed: tcp_close() returned NULL"));
1994-05-24 10:09:53 +00:00
break;
case TCPS_SYN_SENT:
case TCPS_SYN_RECEIVED:
tp->t_flags |= TF_NEEDFIN;
break;
1994-05-24 10:09:53 +00:00
case TCPS_ESTABLISHED:
tp->t_state = TCPS_FIN_WAIT_1;
break;
case TCPS_CLOSE_WAIT:
tp->t_state = TCPS_LAST_ACK;
break;
}
if (tp->t_state >= TCPS_FIN_WAIT_2) {
1994-05-24 10:09:53 +00:00
soisdisconnected(tp->t_inpcb->inp_socket);
/* Prevent the connection hanging in FIN_WAIT_2 forever. */
if (tp->t_state == TCPS_FIN_WAIT_2) {
int timeout;
timeout = (tcp_fast_finwait2_recycle) ?
tcp_finwait2_timeout : TP_MAXIDLE(tp);
tcp_timer_activate(tp, TT_2MSL, timeout);
}
}
1994-05-24 10:09:53 +00:00
}
#ifdef DDB
static void
db_print_indent(int indent)
{
int i;
for (i = 0; i < indent; i++)
db_printf(" ");
}
static void
db_print_tstate(int t_state)
{
switch (t_state) {
case TCPS_CLOSED:
db_printf("TCPS_CLOSED");
return;
case TCPS_LISTEN:
db_printf("TCPS_LISTEN");
return;
case TCPS_SYN_SENT:
db_printf("TCPS_SYN_SENT");
return;
case TCPS_SYN_RECEIVED:
db_printf("TCPS_SYN_RECEIVED");
return;
case TCPS_ESTABLISHED:
db_printf("TCPS_ESTABLISHED");
return;
case TCPS_CLOSE_WAIT:
db_printf("TCPS_CLOSE_WAIT");
return;
case TCPS_FIN_WAIT_1:
db_printf("TCPS_FIN_WAIT_1");
return;
case TCPS_CLOSING:
db_printf("TCPS_CLOSING");
return;
case TCPS_LAST_ACK:
db_printf("TCPS_LAST_ACK");
return;
case TCPS_FIN_WAIT_2:
db_printf("TCPS_FIN_WAIT_2");
return;
case TCPS_TIME_WAIT:
db_printf("TCPS_TIME_WAIT");
return;
default:
db_printf("unknown");
return;
}
}
static void
db_print_tflags(u_int t_flags)
{
int comma;
comma = 0;
if (t_flags & TF_ACKNOW) {
db_printf("%sTF_ACKNOW", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_DELACK) {
db_printf("%sTF_DELACK", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_NODELAY) {
db_printf("%sTF_NODELAY", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_NOOPT) {
db_printf("%sTF_NOOPT", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_SENTFIN) {
db_printf("%sTF_SENTFIN", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_REQ_SCALE) {
db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_RCVD_SCALE) {
db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_REQ_TSTMP) {
db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_RCVD_TSTMP) {
db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_SACK_PERMIT) {
db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_NEEDSYN) {
db_printf("%sTF_NEEDSYN", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_NEEDFIN) {
db_printf("%sTF_NEEDFIN", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_NOPUSH) {
db_printf("%sTF_NOPUSH", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_MORETOCOME) {
db_printf("%sTF_MORETOCOME", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_LQ_OVERFLOW) {
db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_LASTIDLE) {
db_printf("%sTF_LASTIDLE", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_RXWIN0SENT) {
db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_FASTRECOVERY) {
db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
comma = 1;
}
This commit marks the first formal contribution of the "Five New TCP Congestion Control Algorithms for FreeBSD" FreeBSD Foundation funded project. More details about the project are available at: http://caia.swin.edu.au/freebsd/5cc/ - Add a KPI and supporting infrastructure to allow modular congestion control algorithms to be used in the net stack. Algorithms can maintain per-connection state if required, and connections maintain their own algorithm pointer, which allows different connections to concurrently use different algorithms. The TCP_CONGESTION socket option can be used with getsockopt()/setsockopt() to programmatically query or change the congestion control algorithm respectively from within an application at runtime. - Integrate the framework with the TCP stack in as least intrusive a manner as possible. Care was also taken to develop the framework in a way that should allow integration with other congestion aware transport protocols (e.g. SCTP) in the future. The hope is that we will one day be able to share a single set of congestion control algorithm modules between all congestion aware transport protocols. - Introduce a new congestion recovery (TF_CONGRECOVERY) state into the TCP stack and use it to decouple the meaning of recovery from a congestion event and recovery from packet loss (TF_FASTRECOVERY) a la RFC2581. ECN and delay based congestion control protocols don't generally need to recover from packet loss and need a different way to note a congestion recovery episode within the stack. - Remove the net.inet.tcp.newreno sysctl, which simplifies some portions of code and ensures the stack always uses the appropriate mechanisms for recovering from packet loss during a congestion recovery episode. - Extract the NewReno congestion control algorithm from the TCP stack and massage it into module form. NewReno is always built into the kernel and will remain the default algorithm for the forseeable future. Implementations of additional different algorithms will become available in the near future. - Bump __FreeBSD_version to 900025 and note in UPDATING that rebuilding code that relies on the size of "struct tcpcb" is required. Many thanks go to the Cisco University Research Program Fund at Community Foundation Silicon Valley and the FreeBSD Foundation. Their support of our work at the Centre for Advanced Internet Architectures, Swinburne University of Technology is greatly appreciated. In collaboration with: David Hayes <dahayes at swin edu au> and Grenville Armitage <garmitage at swin edu au> Sponsored by: Cisco URP, FreeBSD Foundation Reviewed by: rpaulo Tested by: David Hayes (and many others over the years) MFC after: 3 months
2010-11-12 06:41:55 +00:00
if (t_flags & TF_CONGRECOVERY) {
db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_WASFRECOVERY) {
db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_SIGNATURE) {
db_printf("%sTF_SIGNATURE", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_FORCEDATA) {
db_printf("%sTF_FORCEDATA", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_TSO) {
db_printf("%sTF_TSO", comma ? ", " : "");
comma = 1;
}
if (t_flags & TF_ECN_PERMIT) {
db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
comma = 1;
}
}
static void
db_print_toobflags(char t_oobflags)
{
int comma;
comma = 0;
if (t_oobflags & TCPOOB_HAVEDATA) {
db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
comma = 1;
}
if (t_oobflags & TCPOOB_HADDATA) {
db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
comma = 1;
}
}
static void
db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
{
db_print_indent(indent);
db_printf("%s at %p\n", name, tp);
indent += 2;
db_print_indent(indent);
db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
db_print_indent(indent);
db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
&tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
db_print_indent(indent);
db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
&tp->t_timers->tt_delack, tp->t_inpcb);
db_print_indent(indent);
db_printf("t_state: %d (", tp->t_state);
db_print_tstate(tp->t_state);
db_printf(")\n");
db_print_indent(indent);
db_printf("t_flags: 0x%x (", tp->t_flags);
db_print_tflags(tp->t_flags);
db_printf(")\n");
db_print_indent(indent);
db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
tp->snd_una, tp->snd_max, tp->snd_nxt);
db_print_indent(indent);
db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
tp->snd_up, tp->snd_wl1, tp->snd_wl2);
db_print_indent(indent);
db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
tp->iss, tp->irs, tp->rcv_nxt);
db_print_indent(indent);
db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n",
tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
db_print_indent(indent);
db_printf("snd_wnd: %lu snd_cwnd: %lu\n",
tp->snd_wnd, tp->snd_cwnd);
db_print_indent(indent);
db_printf("snd_ssthresh: %lu snd_recover: "
"0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
db_print_indent(indent);
db_printf("t_maxopd: %u t_rcvtime: %u t_startime: %u\n",
tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
db_print_indent(indent);
db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
tp->t_rtttime, tp->t_rtseq);
db_print_indent(indent);
db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
db_print_indent(indent);
db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
"t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
tp->t_rttbest);
db_print_indent(indent);
db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n",
tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
db_print_indent(indent);
db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
db_print_toobflags(tp->t_oobflags);
db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
db_print_indent(indent);
db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
db_print_indent(indent);
db_printf("ts_recent: %u ts_recent_age: %u\n",
tp->ts_recent, tp->ts_recent_age);
db_print_indent(indent);
db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
"%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
db_print_indent(indent);
db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x "
"t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
tp->snd_recover_prev, tp->t_badrxtwin);
db_print_indent(indent);
db_printf("snd_numholes: %d snd_holes first: %p\n",
tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
db_print_indent(indent);
db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
"0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
/* Skip sackblks, sackhint. */
db_print_indent(indent);
db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
}
DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
{
struct tcpcb *tp;
if (!have_addr) {
db_printf("usage: show tcpcb <addr>\n");
return;
}
tp = (struct tcpcb *)addr;
db_print_tcpcb(tp, "tcpcb", 0);
}
#endif