From 51b9a31c42edcd089f5b229633477ab5128faf03 Mon Sep 17 00:00:00 2001
From: Jon Paul Maloy <jon.maloy@ericsson.com>
Date: Sat, 19 Nov 2016 14:47:07 -0500
Subject: tipc: eliminate obsolete socket locking policy description
The comment block in socket.c describing the locking policy is
obsolete, and does not reflect current reality. We remove it in this
commit.
Since the current locking policy is much simpler and follows a
mainstream approach, we see no need to add a new description.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/tipc/socket.c | 48 +-----------------------------------------------
1 file changed, 1 insertion(+), 47 deletions(-)
(limited to 'net/tipc/socket.c')
diff --git a/net/tipc/socket.c b/net/tipc/socket.c
index f9f5f3c3dab5..db32777ab591 100644
--- a/net/tipc/socket.c
+++ b/net/tipc/socket.c
@@ -1,7 +1,7 @@
/*
* net/tipc/socket.c: TIPC socket API
*
- * Copyright (c) 2001-2007, 2012-2015, Ericsson AB
+ * Copyright (c) 2001-2007, 2012-2016, Ericsson AB
* Copyright (c) 2004-2008, 2010-2013, Wind River Systems
* All rights reserved.
*
@@ -129,54 +129,8 @@ static const struct proto_ops packet_ops;
static const struct proto_ops stream_ops;
static const struct proto_ops msg_ops;
static struct proto tipc_proto;
-
static const struct rhashtable_params tsk_rht_params;
-/*
- * Revised TIPC socket locking policy:
- *
- * Most socket operations take the standard socket lock when they start
- * and hold it until they finish (or until they need to sleep). Acquiring
- * this lock grants the owner exclusive access to the fields of the socket
- * data structures, with the exception of the backlog queue. A few socket
- * operations can be done without taking the socket lock because they only
- * read socket information that never changes during the life of the socket.
- *
- * Socket operations may acquire the lock for the associated TIPC port if they
- * need to perform an operation on the port. If any routine needs to acquire
- * both the socket lock and the port lock it must take the socket lock first
- * to avoid the risk of deadlock.
- *
- * The dispatcher handling incoming messages cannot grab the socket lock in
- * the standard fashion, since invoked it runs at the BH level and cannot block.
- * Instead, it checks to see if the socket lock is currently owned by someone,
- * and either handles the message itself or adds it to the socket's backlog
- * queue; in the latter case the queued message is processed once the process
- * owning the socket lock releases it.
- *
- * NOTE: Releasing the socket lock while an operation is sleeping overcomes
- * the problem of a blocked socket operation preventing any other operations
- * from occurring. However, applications must be careful if they have
- * multiple threads trying to send (or receive) on the same socket, as these
- * operations might interfere with each other. For example, doing a connect
- * and a receive at the same time might allow the receive to consume the
- * ACK message meant for the connect. While additional work could be done
- * to try and overcome this, it doesn't seem to be worthwhile at the present.
- *
- * NOTE: Releasing the socket lock while an operation is sleeping also ensures
- * that another operation that must be performed in a non-blocking manner is
- * not delayed for very long because the lock has already been taken.
- *
- * NOTE: This code assumes that certain fields of a port/socket pair are
- * constant over its lifetime; such fields can be examined without taking
- * the socket lock and/or port lock, and do not need to be re-read even
- * after resuming processing after waiting. These fields include:
- * - socket type
- * - pointer to socket sk structure (aka tipc_sock structure)
- * - pointer to port structure
- * - port reference
- */
-
static u32 tsk_own_node(struct tipc_sock *tsk)
{
return msg_prevnode(&tsk->phdr);
--
cgit v1.2.3
From 6998cc6ec23740347670da13186d2979c5401903 Mon Sep 17 00:00:00 2001
From: Jon Paul Maloy <jon.maloy@ericsson.com>
Date: Thu, 24 Nov 2016 18:47:07 -0500
Subject: tipc: resolve connection flow control compatibility problem
In commit 10724cc7bb78 ("tipc: redesign connection-level flow control")
we replaced the previous message based flow control with one based on
1k blocks. In order to ensure backwards compatibility the mechanism
falls back to using message as base unit when it senses that the peer
doesn't support the new algorithm. The default flow control window,
i.e., how many units can be sent before the sender blocks and waits
for an acknowledge (aka advertisement) is 512. This was tested against
the previous version, which uses an acknowledge frequency of on ack per
256 received message, and found to work fine.
However, we missed the fact that versions older than Linux 3.15 use an
acknowledge frequency of 512, which is exactly the limit where a 4.6+
sender will stop and wait for acknowledge. This would also work fine if
it weren't for the fact that if the first sent message on a 4.6+ server
side is an empty SYNACK, this one is also is counted as a sent message,
while it is not counted as a received message on a legacy 3.15-receiver.
This leads to the sender always being one step ahead of the receiver, a
scenario causing the sender to block after 512 sent messages, while the
receiver only has registered 511 read messages. Hence, the legacy
receiver is not trigged to send an acknowledge, with a permanently
blocked sender as result.
We solve this deadlock by simply allowing the sender to send one more
message before it blocks, i.e., by a making minimal change to the
condition used for determining connection congestion.
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---
net/tipc/socket.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
(limited to 'net/tipc/socket.c')
diff --git a/net/tipc/socket.c b/net/tipc/socket.c
index db32777ab591..41f013888f07 100644
--- a/net/tipc/socket.c
+++ b/net/tipc/socket.c
@@ -186,7 +186,7 @@ static struct tipc_sock *tipc_sk(const struct sock *sk)
static bool tsk_conn_cong(struct tipc_sock *tsk)
{
- return tsk->snt_unacked >= tsk->snd_win;
+ return tsk->snt_unacked > tsk->snd_win;
}
/* tsk_blocks(): translate a buffer size in bytes to number of
--
cgit v1.2.3