2009-08-21 12:28:31 +00:00
|
|
|
/*
|
2018-07-24 03:51:21 +00:00
|
|
|
* Copyright (c) 2006, 2018 Oracle and/or its affiliates. All rights reserved.
|
2009-08-21 12:28:31 +00:00
|
|
|
*
|
|
|
|
* This software is available to you under a choice of one of two
|
|
|
|
* licenses. You may choose to be licensed under the terms of the GNU
|
|
|
|
* General Public License (GPL) Version 2, available from the file
|
|
|
|
* COPYING in the main directory of this source tree, or the
|
|
|
|
* OpenIB.org BSD license below:
|
|
|
|
*
|
|
|
|
* Redistribution and use in source and binary forms, with or
|
|
|
|
* without modification, are permitted provided that the following
|
|
|
|
* conditions are met:
|
|
|
|
*
|
|
|
|
* - Redistributions of source code must retain the above
|
|
|
|
* copyright notice, this list of conditions and the following
|
|
|
|
* disclaimer.
|
|
|
|
*
|
|
|
|
* - 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.
|
|
|
|
*
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
|
|
|
|
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
|
|
|
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
|
|
|
|
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
|
|
|
|
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
|
|
|
|
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
|
|
|
|
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
|
|
* SOFTWARE.
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
#include <linux/kernel.h>
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 08:04:11 +00:00
|
|
|
#include <linux/gfp.h>
|
2009-08-21 12:28:31 +00:00
|
|
|
#include <linux/in.h>
|
|
|
|
#include <net/tcp.h>
|
|
|
|
|
|
|
|
#include "rds.h"
|
|
|
|
#include "tcp.h"
|
|
|
|
|
2020-05-28 05:12:25 +00:00
|
|
|
void rds_tcp_keepalive(struct socket *sock)
|
2015-05-05 19:20:51 +00:00
|
|
|
{
|
|
|
|
/* values below based on xs_udp_default_timeout */
|
|
|
|
int keepidle = 5; /* send a probe 'keepidle' secs after last data */
|
|
|
|
int keepcnt = 5; /* number of unack'ed probes before declaring dead */
|
|
|
|
|
2020-05-28 05:12:15 +00:00
|
|
|
sock_set_keepalive(sock->sk);
|
2020-05-28 05:12:25 +00:00
|
|
|
tcp_sock_set_keepcnt(sock->sk, keepcnt);
|
2020-05-28 05:12:23 +00:00
|
|
|
tcp_sock_set_keepidle(sock->sk, keepidle);
|
2015-05-05 19:20:51 +00:00
|
|
|
/* KEEPINTVL is the interval between successive probes. We follow
|
|
|
|
* the model in xs_tcp_finish_connecting() and re-use keepidle.
|
|
|
|
*/
|
2020-05-28 05:12:24 +00:00
|
|
|
tcp_sock_set_keepintvl(sock->sk, keepidle);
|
2015-05-05 19:20:51 +00:00
|
|
|
}
|
|
|
|
|
2016-07-14 10:51:03 +00:00
|
|
|
/* rds_tcp_accept_one_path(): if accepting on cp_index > 0, make sure the
|
|
|
|
* client's ipaddr < server's ipaddr. Otherwise, close the accepted
|
|
|
|
* socket and force a reconneect from smaller -> larger ip addr. The reason
|
|
|
|
* we special case cp_index 0 is to allow the rds probe ping itself to itself
|
|
|
|
* get through efficiently.
|
|
|
|
* Since reconnects are only initiated from the node with the numerically
|
|
|
|
* smaller ip address, we recycle conns in RDS_CONN_ERROR on the passive side
|
|
|
|
* by moving them to CONNECTING in this function.
|
|
|
|
*/
|
2016-12-05 00:41:29 +00:00
|
|
|
static
|
2016-07-14 10:51:03 +00:00
|
|
|
struct rds_tcp_connection *rds_tcp_accept_one_path(struct rds_connection *conn)
|
|
|
|
{
|
|
|
|
int i;
|
RDS: TCP: Force every connection to be initiated by numerically smaller IP address
When 2 RDS peers initiate an RDS-TCP connection simultaneously,
there is a potential for "duelling syns" on either/both sides.
See commit 241b271952eb ("RDS-TCP: Reset tcp callbacks if re-using an
outgoing socket in rds_tcp_accept_one()") for a description of this
condition, and the arbitration logic which ensures that the
numerically large IP address in the TCP connection is bound to the
RDS_TCP_PORT ("canonical ordering").
The rds_connection should not be marked as RDS_CONN_UP until the
arbitration logic has converged for the following reason. The sender
may start transmitting RDS datagrams as soon as RDS_CONN_UP is set,
and since the sender removes all datagrams from the rds_connection's
cp_retrans queue based on TCP acks. If the TCP ack was sent from
a tcp socket that got reset as part of duel aribitration (but
before data was delivered to the receivers RDS socket layer),
the sender may end up prematurely freeing the datagram, and
the datagram is no longer reliably deliverable.
This patch remedies that condition by making sure that, upon
receipt of 3WH completion state change notification of TCP_ESTABLISHED
in rds_tcp_state_change, we mark the rds_connection as RDS_CONN_UP
if, and only if, the IP addresses and ports for the connection are
canonically ordered. In all other cases, rds_tcp_state_change will
force an rds_conn_path_drop(), and rds_queue_reconnect() on
both peers will restart the connection to ensure canonical ordering.
A side-effect of enforcing this condition in rds_tcp_state_change()
is that rds_tcp_accept_one_path() can now be refactored for simplicity.
It is also no longer possible to encounter an RDS_CONN_UP connection in
the arbitration logic in rds_tcp_accept_one().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-16 21:29:50 +00:00
|
|
|
int npaths = max_t(int, 1, conn->c_npaths);
|
2016-07-14 10:51:03 +00:00
|
|
|
|
RDS: TCP: Force every connection to be initiated by numerically smaller IP address
When 2 RDS peers initiate an RDS-TCP connection simultaneously,
there is a potential for "duelling syns" on either/both sides.
See commit 241b271952eb ("RDS-TCP: Reset tcp callbacks if re-using an
outgoing socket in rds_tcp_accept_one()") for a description of this
condition, and the arbitration logic which ensures that the
numerically large IP address in the TCP connection is bound to the
RDS_TCP_PORT ("canonical ordering").
The rds_connection should not be marked as RDS_CONN_UP until the
arbitration logic has converged for the following reason. The sender
may start transmitting RDS datagrams as soon as RDS_CONN_UP is set,
and since the sender removes all datagrams from the rds_connection's
cp_retrans queue based on TCP acks. If the TCP ack was sent from
a tcp socket that got reset as part of duel aribitration (but
before data was delivered to the receivers RDS socket layer),
the sender may end up prematurely freeing the datagram, and
the datagram is no longer reliably deliverable.
This patch remedies that condition by making sure that, upon
receipt of 3WH completion state change notification of TCP_ESTABLISHED
in rds_tcp_state_change, we mark the rds_connection as RDS_CONN_UP
if, and only if, the IP addresses and ports for the connection are
canonically ordered. In all other cases, rds_tcp_state_change will
force an rds_conn_path_drop(), and rds_queue_reconnect() on
both peers will restart the connection to ensure canonical ordering.
A side-effect of enforcing this condition in rds_tcp_state_change()
is that rds_tcp_accept_one_path() can now be refactored for simplicity.
It is also no longer possible to encounter an RDS_CONN_UP connection in
the arbitration logic in rds_tcp_accept_one().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-16 21:29:50 +00:00
|
|
|
/* for mprds, all paths MUST be initiated by the peer
|
2016-07-14 10:51:03 +00:00
|
|
|
* with the smaller address.
|
|
|
|
*/
|
2018-07-24 03:51:21 +00:00
|
|
|
if (rds_addr_cmp(&conn->c_faddr, &conn->c_laddr) >= 0) {
|
RDS: TCP: Force every connection to be initiated by numerically smaller IP address
When 2 RDS peers initiate an RDS-TCP connection simultaneously,
there is a potential for "duelling syns" on either/both sides.
See commit 241b271952eb ("RDS-TCP: Reset tcp callbacks if re-using an
outgoing socket in rds_tcp_accept_one()") for a description of this
condition, and the arbitration logic which ensures that the
numerically large IP address in the TCP connection is bound to the
RDS_TCP_PORT ("canonical ordering").
The rds_connection should not be marked as RDS_CONN_UP until the
arbitration logic has converged for the following reason. The sender
may start transmitting RDS datagrams as soon as RDS_CONN_UP is set,
and since the sender removes all datagrams from the rds_connection's
cp_retrans queue based on TCP acks. If the TCP ack was sent from
a tcp socket that got reset as part of duel aribitration (but
before data was delivered to the receivers RDS socket layer),
the sender may end up prematurely freeing the datagram, and
the datagram is no longer reliably deliverable.
This patch remedies that condition by making sure that, upon
receipt of 3WH completion state change notification of TCP_ESTABLISHED
in rds_tcp_state_change, we mark the rds_connection as RDS_CONN_UP
if, and only if, the IP addresses and ports for the connection are
canonically ordered. In all other cases, rds_tcp_state_change will
force an rds_conn_path_drop(), and rds_queue_reconnect() on
both peers will restart the connection to ensure canonical ordering.
A side-effect of enforcing this condition in rds_tcp_state_change()
is that rds_tcp_accept_one_path() can now be refactored for simplicity.
It is also no longer possible to encounter an RDS_CONN_UP connection in
the arbitration logic in rds_tcp_accept_one().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-16 21:29:50 +00:00
|
|
|
/* Make sure we initiate at least one path if this
|
|
|
|
* has not already been done; rds_start_mprds() will
|
|
|
|
* take care of additional paths, if necessary.
|
|
|
|
*/
|
|
|
|
if (npaths == 1)
|
|
|
|
rds_conn_path_connect_if_down(&conn->c_path[0]);
|
2016-07-14 10:51:03 +00:00
|
|
|
return NULL;
|
RDS: TCP: Force every connection to be initiated by numerically smaller IP address
When 2 RDS peers initiate an RDS-TCP connection simultaneously,
there is a potential for "duelling syns" on either/both sides.
See commit 241b271952eb ("RDS-TCP: Reset tcp callbacks if re-using an
outgoing socket in rds_tcp_accept_one()") for a description of this
condition, and the arbitration logic which ensures that the
numerically large IP address in the TCP connection is bound to the
RDS_TCP_PORT ("canonical ordering").
The rds_connection should not be marked as RDS_CONN_UP until the
arbitration logic has converged for the following reason. The sender
may start transmitting RDS datagrams as soon as RDS_CONN_UP is set,
and since the sender removes all datagrams from the rds_connection's
cp_retrans queue based on TCP acks. If the TCP ack was sent from
a tcp socket that got reset as part of duel aribitration (but
before data was delivered to the receivers RDS socket layer),
the sender may end up prematurely freeing the datagram, and
the datagram is no longer reliably deliverable.
This patch remedies that condition by making sure that, upon
receipt of 3WH completion state change notification of TCP_ESTABLISHED
in rds_tcp_state_change, we mark the rds_connection as RDS_CONN_UP
if, and only if, the IP addresses and ports for the connection are
canonically ordered. In all other cases, rds_tcp_state_change will
force an rds_conn_path_drop(), and rds_queue_reconnect() on
both peers will restart the connection to ensure canonical ordering.
A side-effect of enforcing this condition in rds_tcp_state_change()
is that rds_tcp_accept_one_path() can now be refactored for simplicity.
It is also no longer possible to encounter an RDS_CONN_UP connection in
the arbitration logic in rds_tcp_accept_one().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-16 21:29:50 +00:00
|
|
|
}
|
2016-07-14 10:51:03 +00:00
|
|
|
|
2016-11-04 17:04:12 +00:00
|
|
|
for (i = 0; i < npaths; i++) {
|
2016-07-14 10:51:03 +00:00
|
|
|
struct rds_conn_path *cp = &conn->c_path[i];
|
|
|
|
|
|
|
|
if (rds_conn_path_transition(cp, RDS_CONN_DOWN,
|
|
|
|
RDS_CONN_CONNECTING) ||
|
|
|
|
rds_conn_path_transition(cp, RDS_CONN_ERROR,
|
|
|
|
RDS_CONN_CONNECTING)) {
|
|
|
|
return cp->cp_transport_data;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2015-08-05 05:43:26 +00:00
|
|
|
int rds_tcp_accept_one(struct socket *sock)
|
2009-08-21 12:28:31 +00:00
|
|
|
{
|
|
|
|
struct socket *new_sock = NULL;
|
|
|
|
struct rds_connection *conn;
|
|
|
|
int ret;
|
|
|
|
struct inet_sock *inet;
|
2016-05-02 18:24:52 +00:00
|
|
|
struct rds_tcp_connection *rs_tcp = NULL;
|
|
|
|
int conn_state;
|
2016-06-30 23:11:14 +00:00
|
|
|
struct rds_conn_path *cp;
|
2018-07-24 03:51:22 +00:00
|
|
|
struct in6_addr *my_addr, *peer_addr;
|
2018-07-31 05:48:42 +00:00
|
|
|
#if !IS_ENABLED(CONFIG_IPV6)
|
|
|
|
struct in6_addr saddr, daddr;
|
|
|
|
#endif
|
|
|
|
int dev_if = 0;
|
2009-08-21 12:28:31 +00:00
|
|
|
|
2016-05-18 17:06:23 +00:00
|
|
|
if (!sock) /* module unload or netns delete in progress */
|
|
|
|
return -ENETUNREACH;
|
|
|
|
|
2017-07-06 15:15:06 +00:00
|
|
|
ret = sock_create_lite(sock->sk->sk_family,
|
2015-08-05 05:43:25 +00:00
|
|
|
sock->sk->sk_type, sock->sk->sk_protocol,
|
|
|
|
&new_sock);
|
2009-08-21 12:28:31 +00:00
|
|
|
if (ret)
|
|
|
|
goto out;
|
|
|
|
|
2017-03-09 08:09:05 +00:00
|
|
|
ret = sock->ops->accept(sock, new_sock, O_NONBLOCK, true);
|
2009-08-21 12:28:31 +00:00
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
2018-03-02 05:07:18 +00:00
|
|
|
/* sock_create_lite() does not get a hold on the owner module so we
|
|
|
|
* need to do it here. Note that sock_release() uses sock->ops to
|
|
|
|
* determine if it needs to decrement the reference count. So set
|
|
|
|
* sock->ops after calling accept() in case that fails. And there's
|
|
|
|
* no need to do try_module_get() as the listener should have a hold
|
|
|
|
* already.
|
|
|
|
*/
|
|
|
|
new_sock->ops = sock->ops;
|
|
|
|
__module_get(new_sock->ops->owner);
|
|
|
|
|
2020-05-28 05:12:25 +00:00
|
|
|
rds_tcp_keepalive(new_sock);
|
2009-08-21 12:28:31 +00:00
|
|
|
rds_tcp_tune(new_sock);
|
|
|
|
|
|
|
|
inet = inet_sk(new_sock->sk);
|
|
|
|
|
2018-07-31 05:48:42 +00:00
|
|
|
#if IS_ENABLED(CONFIG_IPV6)
|
2018-07-24 03:51:22 +00:00
|
|
|
my_addr = &new_sock->sk->sk_v6_rcv_saddr;
|
|
|
|
peer_addr = &new_sock->sk->sk_v6_daddr;
|
2018-07-31 05:48:42 +00:00
|
|
|
#else
|
|
|
|
ipv6_addr_set_v4mapped(inet->inet_saddr, &saddr);
|
|
|
|
ipv6_addr_set_v4mapped(inet->inet_daddr, &daddr);
|
|
|
|
my_addr = &saddr;
|
|
|
|
peer_addr = &daddr;
|
|
|
|
#endif
|
|
|
|
rdsdebug("accepted family %d tcp %pI6c:%u -> %pI6c:%u\n",
|
|
|
|
sock->sk->sk_family,
|
2018-07-24 03:51:22 +00:00
|
|
|
my_addr, ntohs(inet->inet_sport),
|
|
|
|
peer_addr, ntohs(inet->inet_dport));
|
2009-08-21 12:28:31 +00:00
|
|
|
|
2018-07-31 05:48:42 +00:00
|
|
|
#if IS_ENABLED(CONFIG_IPV6)
|
2018-07-24 03:51:22 +00:00
|
|
|
/* sk_bound_dev_if is not set if the peer address is not link local
|
|
|
|
* address. In this case, it happens that mcast_oif is set. So
|
|
|
|
* just use it.
|
|
|
|
*/
|
|
|
|
if ((ipv6_addr_type(my_addr) & IPV6_ADDR_LINKLOCAL) &&
|
|
|
|
!(ipv6_addr_type(peer_addr) & IPV6_ADDR_LINKLOCAL)) {
|
|
|
|
struct ipv6_pinfo *inet6;
|
|
|
|
|
|
|
|
inet6 = inet6_sk(new_sock->sk);
|
|
|
|
dev_if = inet6->mcast_oif;
|
|
|
|
} else {
|
|
|
|
dev_if = new_sock->sk->sk_bound_dev_if;
|
|
|
|
}
|
2018-07-31 05:48:42 +00:00
|
|
|
#endif
|
|
|
|
|
2021-05-21 18:08:06 +00:00
|
|
|
if (!rds_tcp_laddr_check(sock_net(sock->sk), peer_addr, dev_if)) {
|
|
|
|
/* local address connection is only allowed via loopback */
|
|
|
|
ret = -EOPNOTSUPP;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2015-08-05 05:43:25 +00:00
|
|
|
conn = rds_conn_create(sock_net(sock->sk),
|
2018-07-31 05:48:42 +00:00
|
|
|
my_addr, peer_addr,
|
2018-10-24 03:21:14 +00:00
|
|
|
&rds_tcp_transport, 0, GFP_KERNEL, dev_if);
|
2018-07-24 03:51:21 +00:00
|
|
|
|
2009-08-21 12:28:31 +00:00
|
|
|
if (IS_ERR(conn)) {
|
|
|
|
ret = PTR_ERR(conn);
|
|
|
|
goto out;
|
|
|
|
}
|
2015-05-05 19:20:51 +00:00
|
|
|
/* An incoming SYN request came in, and TCP just accepted it.
|
|
|
|
*
|
|
|
|
* If the client reboots, this conn will need to be cleaned up.
|
|
|
|
* rds_tcp_state_change() will do that cleanup
|
|
|
|
*/
|
2016-07-14 10:51:03 +00:00
|
|
|
rs_tcp = rds_tcp_accept_one_path(conn);
|
|
|
|
if (!rs_tcp)
|
|
|
|
goto rst_nsk;
|
2016-06-30 23:11:12 +00:00
|
|
|
mutex_lock(&rs_tcp->t_conn_path_lock);
|
2016-07-14 10:51:03 +00:00
|
|
|
cp = rs_tcp->t_cpath;
|
|
|
|
conn_state = rds_conn_path_state(cp);
|
RDS: TCP: Force every connection to be initiated by numerically smaller IP address
When 2 RDS peers initiate an RDS-TCP connection simultaneously,
there is a potential for "duelling syns" on either/both sides.
See commit 241b271952eb ("RDS-TCP: Reset tcp callbacks if re-using an
outgoing socket in rds_tcp_accept_one()") for a description of this
condition, and the arbitration logic which ensures that the
numerically large IP address in the TCP connection is bound to the
RDS_TCP_PORT ("canonical ordering").
The rds_connection should not be marked as RDS_CONN_UP until the
arbitration logic has converged for the following reason. The sender
may start transmitting RDS datagrams as soon as RDS_CONN_UP is set,
and since the sender removes all datagrams from the rds_connection's
cp_retrans queue based on TCP acks. If the TCP ack was sent from
a tcp socket that got reset as part of duel aribitration (but
before data was delivered to the receivers RDS socket layer),
the sender may end up prematurely freeing the datagram, and
the datagram is no longer reliably deliverable.
This patch remedies that condition by making sure that, upon
receipt of 3WH completion state change notification of TCP_ESTABLISHED
in rds_tcp_state_change, we mark the rds_connection as RDS_CONN_UP
if, and only if, the IP addresses and ports for the connection are
canonically ordered. In all other cases, rds_tcp_state_change will
force an rds_conn_path_drop(), and rds_queue_reconnect() on
both peers will restart the connection to ensure canonical ordering.
A side-effect of enforcing this condition in rds_tcp_state_change()
is that rds_tcp_accept_one_path() can now be refactored for simplicity.
It is also no longer possible to encounter an RDS_CONN_UP connection in
the arbitration logic in rds_tcp_accept_one().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-11-16 21:29:50 +00:00
|
|
|
WARN_ON(conn_state == RDS_CONN_UP);
|
|
|
|
if (conn_state != RDS_CONN_CONNECTING && conn_state != RDS_CONN_ERROR)
|
2016-05-02 18:24:52 +00:00
|
|
|
goto rst_nsk;
|
2016-05-02 18:24:51 +00:00
|
|
|
if (rs_tcp->t_sock) {
|
2017-06-15 18:28:53 +00:00
|
|
|
/* Duelling SYN has been handled in rds_tcp_accept_one() */
|
|
|
|
rds_tcp_reset_callbacks(new_sock, cp);
|
|
|
|
/* rds_connect_path_complete() marks RDS_CONN_UP */
|
|
|
|
rds_connect_path_complete(cp, RDS_CONN_RESETTING);
|
2016-06-04 20:59:58 +00:00
|
|
|
} else {
|
2016-06-30 23:11:14 +00:00
|
|
|
rds_tcp_set_callbacks(new_sock, cp);
|
|
|
|
rds_connect_path_complete(cp, RDS_CONN_CONNECTING);
|
2016-05-02 18:24:51 +00:00
|
|
|
}
|
2009-08-21 12:28:31 +00:00
|
|
|
new_sock = NULL;
|
|
|
|
ret = 0;
|
2017-06-21 20:40:12 +00:00
|
|
|
if (conn->c_npaths == 0)
|
|
|
|
rds_send_ping(cp->cp_conn, cp->cp_index);
|
2016-05-02 18:24:52 +00:00
|
|
|
goto out;
|
|
|
|
rst_nsk:
|
2017-06-15 18:28:55 +00:00
|
|
|
/* reset the newly returned accept sock and bail.
|
|
|
|
* It is safe to set linger on new_sock because the RDS connection
|
|
|
|
* has not been brought up on new_sock, so no RDS-level data could
|
|
|
|
* be pending on it. By setting linger, we achieve the side-effect
|
|
|
|
* of avoiding TIME_WAIT state on new_sock.
|
|
|
|
*/
|
2020-05-28 05:12:10 +00:00
|
|
|
sock_no_linger(new_sock->sk);
|
2016-06-04 20:59:58 +00:00
|
|
|
kernel_sock_shutdown(new_sock, SHUT_RDWR);
|
2016-05-02 18:24:52 +00:00
|
|
|
ret = 0;
|
2009-08-21 12:28:31 +00:00
|
|
|
out:
|
2016-05-02 18:24:52 +00:00
|
|
|
if (rs_tcp)
|
2016-06-30 23:11:12 +00:00
|
|
|
mutex_unlock(&rs_tcp->t_conn_path_lock);
|
2009-08-21 12:28:31 +00:00
|
|
|
if (new_sock)
|
|
|
|
sock_release(new_sock);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2014-04-11 20:15:36 +00:00
|
|
|
void rds_tcp_listen_data_ready(struct sock *sk)
|
2009-08-21 12:28:31 +00:00
|
|
|
{
|
2014-04-11 20:15:36 +00:00
|
|
|
void (*ready)(struct sock *sk);
|
2009-08-21 12:28:31 +00:00
|
|
|
|
|
|
|
rdsdebug("listen data ready sk %p\n", sk);
|
|
|
|
|
2016-05-18 00:44:08 +00:00
|
|
|
read_lock_bh(&sk->sk_callback_lock);
|
2009-08-21 12:28:31 +00:00
|
|
|
ready = sk->sk_user_data;
|
2010-01-12 19:56:44 +00:00
|
|
|
if (!ready) { /* check for teardown race */
|
2009-08-21 12:28:31 +00:00
|
|
|
ready = sk->sk_data_ready;
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* ->sk_data_ready is also called for a newly established child socket
|
|
|
|
* before it has been accepted and the accepter has set up their
|
|
|
|
* data_ready.. we only want to queue listen work for our listening
|
|
|
|
* socket
|
2017-03-04 16:57:35 +00:00
|
|
|
*
|
|
|
|
* (*ready)() may be null if we are racing with netns delete, and
|
|
|
|
* the listen socket is being torn down.
|
2009-08-21 12:28:31 +00:00
|
|
|
*/
|
|
|
|
if (sk->sk_state == TCP_LISTEN)
|
2015-08-05 05:43:26 +00:00
|
|
|
rds_tcp_accept_work(sk);
|
2016-07-14 10:51:01 +00:00
|
|
|
else
|
|
|
|
ready = rds_tcp_listen_sock_def_readable(sock_net(sk));
|
2009-08-21 12:28:31 +00:00
|
|
|
|
|
|
|
out:
|
2016-05-18 00:44:08 +00:00
|
|
|
read_unlock_bh(&sk->sk_callback_lock);
|
2017-03-04 16:57:35 +00:00
|
|
|
if (ready)
|
|
|
|
ready(sk);
|
2009-08-21 12:28:31 +00:00
|
|
|
}
|
|
|
|
|
2018-07-24 03:51:22 +00:00
|
|
|
struct socket *rds_tcp_listen_init(struct net *net, bool isv6)
|
2009-08-21 12:28:31 +00:00
|
|
|
{
|
|
|
|
struct socket *sock = NULL;
|
2018-07-24 03:51:22 +00:00
|
|
|
struct sockaddr_storage ss;
|
|
|
|
struct sockaddr_in6 *sin6;
|
|
|
|
struct sockaddr_in *sin;
|
|
|
|
int addr_len;
|
2009-08-21 12:28:31 +00:00
|
|
|
int ret;
|
|
|
|
|
2018-07-24 03:51:22 +00:00
|
|
|
ret = sock_create_kern(net, isv6 ? PF_INET6 : PF_INET, SOCK_STREAM,
|
|
|
|
IPPROTO_TCP, &sock);
|
|
|
|
if (ret < 0) {
|
|
|
|
rdsdebug("could not create %s listener socket: %d\n",
|
|
|
|
isv6 ? "IPv6" : "IPv4", ret);
|
2009-08-21 12:28:31 +00:00
|
|
|
goto out;
|
2018-07-24 03:51:22 +00:00
|
|
|
}
|
2009-08-21 12:28:31 +00:00
|
|
|
|
2012-04-19 03:39:36 +00:00
|
|
|
sock->sk->sk_reuse = SK_CAN_REUSE;
|
2020-05-28 05:12:19 +00:00
|
|
|
tcp_sock_set_nodelay(sock->sk);
|
2009-08-21 12:28:31 +00:00
|
|
|
|
|
|
|
write_lock_bh(&sock->sk->sk_callback_lock);
|
|
|
|
sock->sk->sk_user_data = sock->sk->sk_data_ready;
|
|
|
|
sock->sk->sk_data_ready = rds_tcp_listen_data_ready;
|
|
|
|
write_unlock_bh(&sock->sk->sk_callback_lock);
|
|
|
|
|
2018-07-24 03:51:22 +00:00
|
|
|
if (isv6) {
|
|
|
|
sin6 = (struct sockaddr_in6 *)&ss;
|
|
|
|
sin6->sin6_family = PF_INET6;
|
|
|
|
sin6->sin6_addr = in6addr_any;
|
|
|
|
sin6->sin6_port = (__force u16)htons(RDS_TCP_PORT);
|
|
|
|
sin6->sin6_scope_id = 0;
|
|
|
|
sin6->sin6_flowinfo = 0;
|
|
|
|
addr_len = sizeof(*sin6);
|
|
|
|
} else {
|
|
|
|
sin = (struct sockaddr_in *)&ss;
|
|
|
|
sin->sin_family = PF_INET;
|
|
|
|
sin->sin_addr.s_addr = INADDR_ANY;
|
|
|
|
sin->sin_port = (__force u16)htons(RDS_TCP_PORT);
|
|
|
|
addr_len = sizeof(*sin);
|
|
|
|
}
|
2009-08-21 12:28:31 +00:00
|
|
|
|
2018-07-24 03:51:22 +00:00
|
|
|
ret = sock->ops->bind(sock, (struct sockaddr *)&ss, addr_len);
|
|
|
|
if (ret < 0) {
|
|
|
|
rdsdebug("could not bind %s listener socket: %d\n",
|
|
|
|
isv6 ? "IPv6" : "IPv4", ret);
|
2009-08-21 12:28:31 +00:00
|
|
|
goto out;
|
2018-07-24 03:51:22 +00:00
|
|
|
}
|
2009-08-21 12:28:31 +00:00
|
|
|
|
|
|
|
ret = sock->ops->listen(sock, 64);
|
|
|
|
if (ret < 0)
|
|
|
|
goto out;
|
|
|
|
|
2015-08-05 05:43:26 +00:00
|
|
|
return sock;
|
2009-08-21 12:28:31 +00:00
|
|
|
out:
|
|
|
|
if (sock)
|
|
|
|
sock_release(sock);
|
2015-08-05 05:43:26 +00:00
|
|
|
return NULL;
|
2009-08-21 12:28:31 +00:00
|
|
|
}
|
|
|
|
|
2017-03-04 16:57:35 +00:00
|
|
|
void rds_tcp_listen_stop(struct socket *sock, struct work_struct *acceptor)
|
2009-08-21 12:28:31 +00:00
|
|
|
{
|
|
|
|
struct sock *sk;
|
|
|
|
|
2010-01-12 19:56:44 +00:00
|
|
|
if (!sock)
|
2009-08-21 12:28:31 +00:00
|
|
|
return;
|
|
|
|
|
|
|
|
sk = sock->sk;
|
|
|
|
|
|
|
|
/* serialize with and prevent further callbacks */
|
|
|
|
lock_sock(sk);
|
|
|
|
write_lock_bh(&sk->sk_callback_lock);
|
|
|
|
if (sk->sk_user_data) {
|
|
|
|
sk->sk_data_ready = sk->sk_user_data;
|
|
|
|
sk->sk_user_data = NULL;
|
|
|
|
}
|
|
|
|
write_unlock_bh(&sk->sk_callback_lock);
|
|
|
|
release_sock(sk);
|
|
|
|
|
|
|
|
/* wait for accepts to stop and close the socket */
|
|
|
|
flush_workqueue(rds_wq);
|
2017-03-04 16:57:35 +00:00
|
|
|
flush_work(acceptor);
|
2009-08-21 12:28:31 +00:00
|
|
|
sock_release(sock);
|
|
|
|
}
|