tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
/*
|
|
|
|
* net/tipc/server.c: TIPC server infrastructure
|
|
|
|
*
|
|
|
|
* Copyright (c) 2012-2013, Wind River Systems
|
2018-02-15 09:40:51 +00:00
|
|
|
* Copyright (c) 2017-2018, Ericsson AB
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* 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.
|
|
|
|
* 3. Neither the names of the copyright holders nor the names of its
|
|
|
|
* contributors may be used to endorse or promote products derived from
|
|
|
|
* this software without specific prior written permission.
|
|
|
|
*
|
|
|
|
* Alternatively, this software may be distributed under the terms of the
|
|
|
|
* GNU General Public License ("GPL") version 2 as published by the Free
|
|
|
|
* Software Foundation.
|
|
|
|
*
|
|
|
|
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
|
|
|
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
|
|
|
|
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
|
|
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
|
|
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
|
|
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
|
|
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
|
|
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
|
|
*/
|
|
|
|
|
2018-02-15 09:40:43 +00:00
|
|
|
#include "subscr.h"
|
2018-02-15 09:40:51 +00:00
|
|
|
#include "topsrv.h"
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
#include "core.h"
|
2015-01-09 07:27:01 +00:00
|
|
|
#include "socket.h"
|
2017-10-13 09:04:17 +00:00
|
|
|
#include "addr.h"
|
|
|
|
#include "msg.h"
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
#include <net/sock.h>
|
tipc: fix netns refcnt leak
When the TIPC module is loaded, we launch a topology server in kernel
space, which in its turn is creating TIPC sockets for communication
with topology server users. Because both the socket's creator and
provider reside in the same module, it is necessary that the TIPC
module's reference count remains zero after the server is started and
the socket created; otherwise it becomes impossible to perform "rmmod"
even on an idle module.
Currently, we achieve this by defining a separate "tipc_proto_kern"
protocol struct, that is used only for kernel space socket allocations.
This structure has the "owner" field set to NULL, which restricts the
module reference count from being be bumped when sk_alloc() for local
sockets is called. Furthermore, we have defined three kernel-specific
functions, tipc_sock_create_local(), tipc_sock_release_local() and
tipc_sock_accept_local(), to avoid the module counter being modified
when module local sockets are created or deleted. This has worked well
until we introduced name space support.
However, after name space support was introduced, we have observed that
a reference count leak occurs, because the netns counter is not
decremented in tipc_sock_delete_local().
This commit remedies this problem. But instead of just modifying
tipc_sock_delete_local(), we eliminate the whole parallel socket
handling infrastructure, and start using the regular sk_create_kern(),
kernel_accept() and sk_release_kernel() calls. Since those functions
manipulate the module counter, we must now compensate for that by
explicitly decrementing the counter after module local sockets are
created, and increment it just before calling sk_release_kernel().
Fixes: a62fbccecd62 ("tipc: make subscriber server support net namespace")
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Jon Maloy <jon.maloy@ericson.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reported-by: Cong Wang <cwang@twopensource.com>
Tested-by: Erik Hugne <erik.hugne@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-18 01:32:57 +00:00
|
|
|
#include <linux/module.h>
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
|
|
|
/* Number of messages to send before rescheduling */
|
|
|
|
#define MAX_SEND_MSG_COUNT 25
|
|
|
|
#define MAX_RECV_MSG_COUNT 25
|
|
|
|
#define CF_CONNECTED 1
|
tipc: fix netns refcnt leak
When the TIPC module is loaded, we launch a topology server in kernel
space, which in its turn is creating TIPC sockets for communication
with topology server users. Because both the socket's creator and
provider reside in the same module, it is necessary that the TIPC
module's reference count remains zero after the server is started and
the socket created; otherwise it becomes impossible to perform "rmmod"
even on an idle module.
Currently, we achieve this by defining a separate "tipc_proto_kern"
protocol struct, that is used only for kernel space socket allocations.
This structure has the "owner" field set to NULL, which restricts the
module reference count from being be bumped when sk_alloc() for local
sockets is called. Furthermore, we have defined three kernel-specific
functions, tipc_sock_create_local(), tipc_sock_release_local() and
tipc_sock_accept_local(), to avoid the module counter being modified
when module local sockets are created or deleted. This has worked well
until we introduced name space support.
However, after name space support was introduced, we have observed that
a reference count leak occurs, because the netns counter is not
decremented in tipc_sock_delete_local().
This commit remedies this problem. But instead of just modifying
tipc_sock_delete_local(), we eliminate the whole parallel socket
handling infrastructure, and start using the regular sk_create_kern(),
kernel_accept() and sk_release_kernel() calls. Since those functions
manipulate the module counter, we must now compensate for that by
explicitly decrementing the counter after module local sockets are
created, and increment it just before calling sk_release_kernel().
Fixes: a62fbccecd62 ("tipc: make subscriber server support net namespace")
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Jon Maloy <jon.maloy@ericson.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reported-by: Cong Wang <cwang@twopensource.com>
Tested-by: Erik Hugne <erik.hugne@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-18 01:32:57 +00:00
|
|
|
#define CF_SERVER 2
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
2018-02-15 09:40:49 +00:00
|
|
|
#define TIPC_SERVER_NAME_LEN 32
|
|
|
|
|
|
|
|
/**
|
2018-02-15 09:40:51 +00:00
|
|
|
* struct tipc_topsrv - TIPC server structure
|
2018-02-15 09:40:49 +00:00
|
|
|
* @conn_idr: identifier set of connection
|
|
|
|
* @idr_lock: protect the connection identifier set
|
|
|
|
* @idr_in_use: amount of allocated identifier entry
|
|
|
|
* @net: network namspace instance
|
2018-09-03 06:08:40 +00:00
|
|
|
* @awork: accept work item
|
2018-02-15 09:40:49 +00:00
|
|
|
* @rcv_wq: receive workqueue
|
|
|
|
* @send_wq: send workqueue
|
|
|
|
* @max_rcvbuf_size: maximum permitted receive message length
|
2018-09-03 06:08:40 +00:00
|
|
|
* @listener: topsrv listener socket
|
2018-02-15 09:40:49 +00:00
|
|
|
* @name: server name
|
|
|
|
*/
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv {
|
2018-02-15 09:40:49 +00:00
|
|
|
struct idr conn_idr;
|
|
|
|
spinlock_t idr_lock; /* for idr list */
|
|
|
|
int idr_in_use;
|
|
|
|
struct net *net;
|
2018-02-15 09:40:50 +00:00
|
|
|
struct work_struct awork;
|
2018-02-15 09:40:49 +00:00
|
|
|
struct workqueue_struct *rcv_wq;
|
|
|
|
struct workqueue_struct *send_wq;
|
|
|
|
int max_rcvbuf_size;
|
2018-02-15 09:40:50 +00:00
|
|
|
struct socket *listener;
|
2018-02-15 09:40:49 +00:00
|
|
|
char name[TIPC_SERVER_NAME_LEN];
|
|
|
|
};
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
|
|
|
/**
|
|
|
|
* struct tipc_conn - TIPC connection structure
|
|
|
|
* @kref: reference counter to connection object
|
|
|
|
* @conid: connection identifier
|
|
|
|
* @sock: socket handler associated with connection
|
|
|
|
* @flags: indicates connection state
|
|
|
|
* @server: pointer to connected server
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
* @sub_list: lsit to all pertaing subscriptions
|
|
|
|
* @sub_lock: lock protecting the subscription list
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
* @rwork: receive work item
|
|
|
|
* @outqueue: pointer to first outbound message in queue
|
2014-01-12 20:48:00 +00:00
|
|
|
* @outqueue_lock: control access to the outqueue
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
* @swork: send work item
|
|
|
|
*/
|
|
|
|
struct tipc_conn {
|
|
|
|
struct kref kref;
|
|
|
|
int conid;
|
|
|
|
struct socket *sock;
|
|
|
|
unsigned long flags;
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *server;
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
struct list_head sub_list;
|
|
|
|
spinlock_t sub_lock; /* for subscription list */
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
struct work_struct rwork;
|
|
|
|
struct list_head outqueue;
|
2018-02-15 09:40:51 +00:00
|
|
|
spinlock_t outqueue_lock; /* for outqueue */
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
struct work_struct swork;
|
|
|
|
};
|
|
|
|
|
|
|
|
/* An entry waiting to be sent */
|
|
|
|
struct outqueue_entry {
|
2018-02-15 09:40:45 +00:00
|
|
|
bool inactive;
|
|
|
|
struct tipc_event evt;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
struct list_head list;
|
|
|
|
};
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_conn_recv_work(struct work_struct *work);
|
|
|
|
static void tipc_conn_send_work(struct work_struct *work);
|
|
|
|
static void tipc_topsrv_kern_evt(struct net *net, struct tipc_event *evt);
|
|
|
|
static void tipc_conn_delete_sub(struct tipc_conn *con, struct tipc_subscr *s);
|
2018-02-15 09:40:49 +00:00
|
|
|
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
static bool connected(struct tipc_conn *con)
|
|
|
|
{
|
|
|
|
return con && test_bit(CF_CONNECTED, &con->flags);
|
|
|
|
}
|
|
|
|
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
static void tipc_conn_kref_release(struct kref *kref)
|
|
|
|
{
|
|
|
|
struct tipc_conn *con = container_of(kref, struct tipc_conn, kref);
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *s = con->server;
|
2018-02-15 09:40:50 +00:00
|
|
|
struct outqueue_entry *e, *safe;
|
tipc: fix netns refcnt leak
When the TIPC module is loaded, we launch a topology server in kernel
space, which in its turn is creating TIPC sockets for communication
with topology server users. Because both the socket's creator and
provider reside in the same module, it is necessary that the TIPC
module's reference count remains zero after the server is started and
the socket created; otherwise it becomes impossible to perform "rmmod"
even on an idle module.
Currently, we achieve this by defining a separate "tipc_proto_kern"
protocol struct, that is used only for kernel space socket allocations.
This structure has the "owner" field set to NULL, which restricts the
module reference count from being be bumped when sk_alloc() for local
sockets is called. Furthermore, we have defined three kernel-specific
functions, tipc_sock_create_local(), tipc_sock_release_local() and
tipc_sock_accept_local(), to avoid the module counter being modified
when module local sockets are created or deleted. This has worked well
until we introduced name space support.
However, after name space support was introduced, we have observed that
a reference count leak occurs, because the netns counter is not
decremented in tipc_sock_delete_local().
This commit remedies this problem. But instead of just modifying
tipc_sock_delete_local(), we eliminate the whole parallel socket
handling infrastructure, and start using the regular sk_create_kern(),
kernel_accept() and sk_release_kernel() calls. Since those functions
manipulate the module counter, we must now compensate for that by
explicitly decrementing the counter after module local sockets are
created, and increment it just before calling sk_release_kernel().
Fixes: a62fbccecd62 ("tipc: make subscriber server support net namespace")
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Reviewed-by: Jon Maloy <jon.maloy@ericson.com>
Reviewed-by: Erik Hugne <erik.hugne@ericsson.com>
Reported-by: Cong Wang <cwang@twopensource.com>
Tested-by: Erik Hugne <erik.hugne@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2015-03-18 01:32:57 +00:00
|
|
|
|
2017-10-13 09:04:17 +00:00
|
|
|
spin_lock_bh(&s->idr_lock);
|
|
|
|
idr_remove(&s->conn_idr, con->conid);
|
|
|
|
s->idr_in_use--;
|
|
|
|
spin_unlock_bh(&s->idr_lock);
|
2018-02-15 09:40:50 +00:00
|
|
|
if (con->sock)
|
|
|
|
sock_release(con->sock);
|
|
|
|
|
|
|
|
spin_lock_bh(&con->outqueue_lock);
|
|
|
|
list_for_each_entry_safe(e, safe, &con->outqueue, list) {
|
|
|
|
list_del(&e->list);
|
|
|
|
kfree(e);
|
|
|
|
}
|
|
|
|
spin_unlock_bh(&con->outqueue_lock);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
kfree(con);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void conn_put(struct tipc_conn *con)
|
|
|
|
{
|
|
|
|
kref_put(&con->kref, tipc_conn_kref_release);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void conn_get(struct tipc_conn *con)
|
|
|
|
{
|
|
|
|
kref_get(&con->kref);
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_conn_close(struct tipc_conn *con)
|
2016-04-12 11:05:21 +00:00
|
|
|
{
|
2018-01-15 16:56:28 +00:00
|
|
|
struct sock *sk = con->sock->sk;
|
|
|
|
bool disconnect = false;
|
2016-04-12 11:05:21 +00:00
|
|
|
|
2018-01-15 16:56:28 +00:00
|
|
|
write_lock_bh(&sk->sk_callback_lock);
|
|
|
|
disconnect = test_and_clear_bit(CF_CONNECTED, &con->flags);
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
|
2018-01-15 16:56:28 +00:00
|
|
|
if (disconnect) {
|
|
|
|
sk->sk_user_data = NULL;
|
2018-02-15 09:40:51 +00:00
|
|
|
tipc_conn_delete_sub(con, NULL);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
2018-01-15 16:56:28 +00:00
|
|
|
write_unlock_bh(&sk->sk_callback_lock);
|
|
|
|
|
|
|
|
/* Handle concurrent calls from sending and receiving threads */
|
|
|
|
if (!disconnect)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* Don't flush pending works, -just let them expire */
|
|
|
|
kernel_sock_shutdown(con->sock, SHUT_RDWR);
|
2018-02-15 09:40:50 +00:00
|
|
|
|
2018-01-15 16:56:28 +00:00
|
|
|
conn_put(con);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static struct tipc_conn *tipc_conn_alloc(struct tipc_topsrv *s)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
|
|
|
struct tipc_conn *con;
|
|
|
|
int ret;
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
con = kzalloc(sizeof(*con), GFP_ATOMIC);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
if (!con)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
|
|
|
|
kref_init(&con->kref);
|
|
|
|
INIT_LIST_HEAD(&con->outqueue);
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
INIT_LIST_HEAD(&con->sub_list);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
spin_lock_init(&con->outqueue_lock);
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
spin_lock_init(&con->sub_lock);
|
2018-02-15 09:40:51 +00:00
|
|
|
INIT_WORK(&con->swork, tipc_conn_send_work);
|
|
|
|
INIT_WORK(&con->rwork, tipc_conn_recv_work);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
|
|
|
spin_lock_bh(&s->idr_lock);
|
|
|
|
ret = idr_alloc(&s->conn_idr, con, 0, 0, GFP_ATOMIC);
|
|
|
|
if (ret < 0) {
|
|
|
|
kfree(con);
|
|
|
|
spin_unlock_bh(&s->idr_lock);
|
|
|
|
return ERR_PTR(-ENOMEM);
|
|
|
|
}
|
|
|
|
con->conid = ret;
|
|
|
|
s->idr_in_use++;
|
|
|
|
spin_unlock_bh(&s->idr_lock);
|
|
|
|
|
|
|
|
set_bit(CF_CONNECTED, &con->flags);
|
|
|
|
con->server = s;
|
|
|
|
|
|
|
|
return con;
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static struct tipc_conn *tipc_conn_lookup(struct tipc_topsrv *s, int conid)
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_conn *con;
|
|
|
|
|
|
|
|
spin_lock_bh(&s->idr_lock);
|
|
|
|
con = idr_find(&s->conn_idr, conid);
|
|
|
|
if (!connected(con) || !kref_get_unless_zero(&con->kref))
|
|
|
|
con = NULL;
|
|
|
|
spin_unlock_bh(&s->idr_lock);
|
|
|
|
return con;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* tipc_conn_delete_sub - delete a specific or all subscriptions
|
|
|
|
* for a given subscriber
|
|
|
|
*/
|
|
|
|
static void tipc_conn_delete_sub(struct tipc_conn *con, struct tipc_subscr *s)
|
|
|
|
{
|
|
|
|
struct tipc_net *tn = tipc_net(con->server->net);
|
|
|
|
struct list_head *sub_list = &con->sub_list;
|
|
|
|
struct tipc_subscription *sub, *tmp;
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
|
|
|
|
spin_lock_bh(&con->sub_lock);
|
2018-02-15 09:40:51 +00:00
|
|
|
list_for_each_entry_safe(sub, tmp, sub_list, sub_list) {
|
|
|
|
if (!s || !memcmp(s, &sub->evt.s, sizeof(*s))) {
|
|
|
|
tipc_sub_unsubscribe(sub);
|
|
|
|
atomic_dec(&tn->subscription_count);
|
|
|
|
} else if (s) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
spin_unlock_bh(&con->sub_lock);
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_conn_send_to_sock(struct tipc_conn *con)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct list_head *queue = &con->outqueue;
|
|
|
|
struct tipc_topsrv *srv = con->server;
|
|
|
|
struct outqueue_entry *e;
|
|
|
|
struct tipc_event *evt;
|
|
|
|
struct msghdr msg;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
struct kvec iov;
|
2018-02-15 09:40:51 +00:00
|
|
|
int count = 0;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
int ret;
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
spin_lock_bh(&con->outqueue_lock);
|
|
|
|
|
|
|
|
while (!list_empty(queue)) {
|
|
|
|
e = list_first_entry(queue, struct outqueue_entry, list);
|
|
|
|
evt = &e->evt;
|
|
|
|
spin_unlock_bh(&con->outqueue_lock);
|
|
|
|
|
|
|
|
if (e->inactive)
|
|
|
|
tipc_conn_delete_sub(con, &evt->s);
|
|
|
|
|
|
|
|
memset(&msg, 0, sizeof(msg));
|
|
|
|
msg.msg_flags = MSG_DONTWAIT;
|
|
|
|
iov.iov_base = evt;
|
|
|
|
iov.iov_len = sizeof(*evt);
|
|
|
|
msg.msg_name = NULL;
|
|
|
|
|
|
|
|
if (con->sock) {
|
|
|
|
ret = kernel_sendmsg(con->sock, &msg, &iov,
|
|
|
|
1, sizeof(*evt));
|
|
|
|
if (ret == -EWOULDBLOCK || ret == 0) {
|
|
|
|
cond_resched();
|
|
|
|
return;
|
|
|
|
} else if (ret < 0) {
|
|
|
|
return tipc_conn_close(con);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
tipc_topsrv_kern_evt(srv->net, evt);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Don't starve users filling buffers */
|
|
|
|
if (++count >= MAX_SEND_MSG_COUNT) {
|
|
|
|
cond_resched();
|
|
|
|
count = 0;
|
|
|
|
}
|
|
|
|
spin_lock_bh(&con->outqueue_lock);
|
|
|
|
list_del(&e->list);
|
|
|
|
kfree(e);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
2018-02-15 09:40:51 +00:00
|
|
|
spin_unlock_bh(&con->outqueue_lock);
|
|
|
|
}
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_conn_send_work(struct work_struct *work)
|
|
|
|
{
|
|
|
|
struct tipc_conn *con = container_of(work, struct tipc_conn, swork);
|
|
|
|
|
|
|
|
if (connected(con))
|
|
|
|
tipc_conn_send_to_sock(con);
|
|
|
|
|
|
|
|
conn_put(con);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
|
2018-09-03 08:36:46 +00:00
|
|
|
/* tipc_topsrv_queue_evt() - interrupt level call from a subscription instance
|
|
|
|
* The queued work is launched into tipc_conn_send_work()->tipc_conn_send_to_sock()
|
2018-02-15 09:40:45 +00:00
|
|
|
*/
|
2018-02-15 09:40:51 +00:00
|
|
|
void tipc_topsrv_queue_evt(struct net *net, int conid,
|
|
|
|
u32 event, struct tipc_event *evt)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *srv = tipc_topsrv(net);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
struct outqueue_entry *e;
|
|
|
|
struct tipc_conn *con;
|
|
|
|
|
2018-02-15 09:40:49 +00:00
|
|
|
con = tipc_conn_lookup(srv, conid);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
if (!con)
|
2018-02-15 09:40:45 +00:00
|
|
|
return;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
2018-02-15 09:40:45 +00:00
|
|
|
if (!connected(con))
|
|
|
|
goto err;
|
2017-01-24 12:00:47 +00:00
|
|
|
|
2018-02-15 09:40:45 +00:00
|
|
|
e = kmalloc(sizeof(*e), GFP_ATOMIC);
|
|
|
|
if (!e)
|
|
|
|
goto err;
|
|
|
|
e->inactive = (event == TIPC_SUBSCR_TIMEOUT);
|
|
|
|
memcpy(&e->evt, evt, sizeof(*evt));
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
spin_lock_bh(&con->outqueue_lock);
|
|
|
|
list_add_tail(&e->list, &con->outqueue);
|
|
|
|
spin_unlock_bh(&con->outqueue_lock);
|
|
|
|
|
2018-02-15 09:40:49 +00:00
|
|
|
if (queue_work(srv->send_wq, &con->swork))
|
2018-02-15 09:40:45 +00:00
|
|
|
return;
|
|
|
|
err:
|
|
|
|
conn_put(con);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
/* tipc_conn_write_space - interrupt callback after a sendmsg EAGAIN
|
|
|
|
* Indicates that there now is more space in the send buffer
|
|
|
|
* The queued work is launched into tipc_send_work()->tipc_conn_send_to_sock()
|
|
|
|
*/
|
|
|
|
static void tipc_conn_write_space(struct sock *sk)
|
2017-10-13 09:04:17 +00:00
|
|
|
{
|
|
|
|
struct tipc_conn *con;
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
read_lock_bh(&sk->sk_callback_lock);
|
|
|
|
con = sk->sk_user_data;
|
|
|
|
if (connected(con)) {
|
|
|
|
conn_get(con);
|
|
|
|
if (!queue_work(con->server->send_wq, &con->swork))
|
|
|
|
conn_put(con);
|
|
|
|
}
|
|
|
|
read_unlock_bh(&sk->sk_callback_lock);
|
2017-10-13 09:04:17 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static int tipc_conn_rcv_sub(struct tipc_topsrv *srv,
|
|
|
|
struct tipc_conn *con,
|
|
|
|
struct tipc_subscr *s)
|
2017-10-13 09:04:17 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_net *tn = tipc_net(srv->net);
|
|
|
|
struct tipc_subscription *sub;
|
2017-10-13 09:04:17 +00:00
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
if (tipc_sub_read(s, filter) & TIPC_SUB_CANCEL) {
|
|
|
|
tipc_conn_delete_sub(con, s);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
if (atomic_read(&tn->subscription_count) >= TIPC_MAX_SUBSCR) {
|
|
|
|
pr_warn("Subscription rejected, max (%u)\n", TIPC_MAX_SUBSCR);
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
sub = tipc_sub_subscribe(srv->net, s, con->conid);
|
|
|
|
if (!sub)
|
|
|
|
return -1;
|
|
|
|
atomic_inc(&tn->subscription_count);
|
|
|
|
spin_lock_bh(&con->sub_lock);
|
|
|
|
list_add(&sub->sub_list, &con->sub_list);
|
|
|
|
spin_unlock_bh(&con->sub_lock);
|
|
|
|
return 0;
|
2017-10-13 09:04:17 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static int tipc_conn_rcv_from_sock(struct tipc_conn *con)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *srv = con->server;
|
|
|
|
struct sock *sk = con->sock->sk;
|
|
|
|
struct msghdr msg = {};
|
|
|
|
struct tipc_subscr s;
|
2018-02-15 09:40:45 +00:00
|
|
|
struct kvec iov;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
int ret;
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
iov.iov_base = &s;
|
|
|
|
iov.iov_len = sizeof(s);
|
|
|
|
msg.msg_name = NULL;
|
2018-10-19 23:57:56 +00:00
|
|
|
iov_iter_kvec(&msg.msg_iter, READ, &iov, 1, iov.iov_len);
|
2018-02-15 09:40:51 +00:00
|
|
|
ret = sock_recvmsg(con->sock, &msg, MSG_DONTWAIT);
|
|
|
|
if (ret == -EWOULDBLOCK)
|
|
|
|
return -EWOULDBLOCK;
|
|
|
|
if (ret > 0) {
|
|
|
|
read_lock_bh(&sk->sk_callback_lock);
|
|
|
|
ret = tipc_conn_rcv_sub(srv, con, &s);
|
|
|
|
read_unlock_bh(&sk->sk_callback_lock);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
2018-02-15 09:40:51 +00:00
|
|
|
if (ret < 0)
|
|
|
|
tipc_conn_close(con);
|
|
|
|
|
|
|
|
return ret;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_conn_recv_work(struct work_struct *work)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
|
|
|
struct tipc_conn *con = container_of(work, struct tipc_conn, rwork);
|
|
|
|
int count = 0;
|
|
|
|
|
tipc: eliminate struct tipc_subscriber
It is unnecessary to keep two structures, struct tipc_conn and struct
tipc_subscriber, with a one-to-one relationship and still with different
life cycles. The fact that the two often run in different contexts, and
still may access each other via direct pointers constitutes an additional
hazard, something we have experienced at several occasions, and still
see happening.
We have identified at least two remaining problems that are easier to
fix if we simplify the topology server data structure somewhat.
- When there is a race between a subscription up/down event and a
timeout event, it is fully possible that the former might be delivered
after the latter, leading to confusion for the receiver.
- The function tipc_subcrp_timeout() is executing in interrupt context,
while the following call chain is at least theoretically possible:
tipc_subscrp_timeout()
tipc_subscrp_send_event()
tipc_conn_sendmsg()
conn_put()
tipc_conn_kref_release()
sock_release(sock)
I.e., we end up calling a function that might try to sleep in
interrupt context. To eliminate this, we need to ensure that the
tipc_conn structure and the socket, as well as the subscription
instances, only are deleted in work queue context, i.e., after the
timeout event really has been sent out.
We now remove this unnecessary complexity, by merging data and
functionality of the subscriber structure into struct tipc_conn
and the associated file server.c. We thereafter add a spinlock and
a new 'inactive' state to the subscription structure. Using those,
both problems described above can be easily solved.
Acked-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-15 09:40:44 +00:00
|
|
|
while (connected(con)) {
|
2018-02-15 09:40:51 +00:00
|
|
|
if (tipc_conn_rcv_from_sock(con))
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
break;
|
|
|
|
|
|
|
|
/* Don't flood Rx machine */
|
|
|
|
if (++count >= MAX_RECV_MSG_COUNT) {
|
|
|
|
cond_resched();
|
|
|
|
count = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
conn_put(con);
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
/* tipc_conn_data_ready - interrupt callback indicating the socket has data
|
|
|
|
* The queued work is launched into tipc_recv_work()->tipc_conn_rcv_from_sock()
|
|
|
|
*/
|
|
|
|
static void tipc_conn_data_ready(struct sock *sk)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_conn *con;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
read_lock_bh(&sk->sk_callback_lock);
|
|
|
|
con = sk->sk_user_data;
|
|
|
|
if (connected(con)) {
|
|
|
|
conn_get(con);
|
|
|
|
if (!queue_work(con->server->rcv_wq, &con->rwork))
|
|
|
|
conn_put(con);
|
|
|
|
}
|
|
|
|
read_unlock_bh(&sk->sk_callback_lock);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_topsrv_accept(struct work_struct *work)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *srv = container_of(work, struct tipc_topsrv, awork);
|
2018-02-15 09:40:50 +00:00
|
|
|
struct socket *lsock = srv->listener;
|
|
|
|
struct socket *newsock;
|
|
|
|
struct tipc_conn *con;
|
|
|
|
struct sock *newsk;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
while (1) {
|
|
|
|
ret = kernel_accept(lsock, &newsock, O_NONBLOCK);
|
|
|
|
if (ret < 0)
|
|
|
|
return;
|
2018-02-15 09:40:51 +00:00
|
|
|
con = tipc_conn_alloc(srv);
|
2018-02-15 09:40:50 +00:00
|
|
|
if (IS_ERR(con)) {
|
|
|
|
ret = PTR_ERR(con);
|
|
|
|
sock_release(newsock);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
/* Register callbacks */
|
|
|
|
newsk = newsock->sk;
|
|
|
|
write_lock_bh(&newsk->sk_callback_lock);
|
2018-02-15 09:40:51 +00:00
|
|
|
newsk->sk_data_ready = tipc_conn_data_ready;
|
|
|
|
newsk->sk_write_space = tipc_conn_write_space;
|
2018-02-15 09:40:50 +00:00
|
|
|
newsk->sk_user_data = con;
|
|
|
|
con->sock = newsock;
|
|
|
|
write_unlock_bh(&newsk->sk_callback_lock);
|
|
|
|
|
|
|
|
/* Wake up receive process in case of 'SYN+' message */
|
|
|
|
newsk->sk_data_ready(newsk);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
/* tipc_toprsv_listener_data_ready - interrupt callback with connection request
|
|
|
|
* The queued job is launched into tipc_topsrv_accept()
|
2018-02-15 09:40:50 +00:00
|
|
|
*/
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_topsrv_listener_data_ready(struct sock *sk)
|
2018-02-15 09:40:50 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *srv;
|
2018-02-15 09:40:50 +00:00
|
|
|
|
|
|
|
read_lock_bh(&sk->sk_callback_lock);
|
|
|
|
srv = sk->sk_user_data;
|
|
|
|
if (srv->listener)
|
|
|
|
queue_work(srv->rcv_wq, &srv->awork);
|
|
|
|
read_unlock_bh(&sk->sk_callback_lock);
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static int tipc_topsrv_create_listener(struct tipc_topsrv *srv)
|
2018-02-15 09:40:50 +00:00
|
|
|
{
|
|
|
|
int imp = TIPC_CRITICAL_IMPORTANCE;
|
|
|
|
struct socket *lsock = NULL;
|
|
|
|
struct sockaddr_tipc saddr;
|
|
|
|
struct sock *sk;
|
|
|
|
int rc;
|
|
|
|
|
|
|
|
rc = sock_create_kern(srv->net, AF_TIPC, SOCK_SEQPACKET, 0, &lsock);
|
|
|
|
if (rc < 0)
|
|
|
|
return rc;
|
|
|
|
|
|
|
|
srv->listener = lsock;
|
|
|
|
sk = lsock->sk;
|
|
|
|
write_lock_bh(&sk->sk_callback_lock);
|
2018-02-15 09:40:51 +00:00
|
|
|
sk->sk_data_ready = tipc_topsrv_listener_data_ready;
|
2018-02-15 09:40:50 +00:00
|
|
|
sk->sk_user_data = srv;
|
|
|
|
write_unlock_bh(&sk->sk_callback_lock);
|
|
|
|
|
|
|
|
rc = kernel_setsockopt(lsock, SOL_TIPC, TIPC_IMPORTANCE,
|
|
|
|
(char *)&imp, sizeof(imp));
|
|
|
|
if (rc < 0)
|
|
|
|
goto err;
|
|
|
|
|
|
|
|
saddr.family = AF_TIPC;
|
|
|
|
saddr.addrtype = TIPC_ADDR_NAMESEQ;
|
|
|
|
saddr.addr.nameseq.type = TIPC_TOP_SRV;
|
|
|
|
saddr.addr.nameseq.lower = TIPC_TOP_SRV;
|
|
|
|
saddr.addr.nameseq.upper = TIPC_TOP_SRV;
|
|
|
|
saddr.scope = TIPC_NODE_SCOPE;
|
|
|
|
|
|
|
|
rc = kernel_bind(lsock, (struct sockaddr *)&saddr, sizeof(saddr));
|
|
|
|
if (rc < 0)
|
|
|
|
goto err;
|
|
|
|
rc = kernel_listen(lsock, 0);
|
|
|
|
if (rc < 0)
|
|
|
|
goto err;
|
|
|
|
|
|
|
|
/* As server's listening socket owner and creator is the same module,
|
|
|
|
* we have to decrease TIPC module reference count to guarantee that
|
|
|
|
* it remains zero after the server socket is created, otherwise,
|
|
|
|
* executing "rmmod" command is unable to make TIPC module deleted
|
|
|
|
* after TIPC module is inserted successfully.
|
|
|
|
*
|
|
|
|
* However, the reference count is ever increased twice in
|
|
|
|
* sock_create_kern(): one is to increase the reference count of owner
|
|
|
|
* of TIPC socket's proto_ops struct; another is to increment the
|
|
|
|
* reference count of owner of TIPC proto struct. Therefore, we must
|
|
|
|
* decrement the module reference count twice to ensure that it keeps
|
|
|
|
* zero after server's listening socket is created. Of course, we
|
|
|
|
* must bump the module reference count twice as well before the socket
|
|
|
|
* is closed.
|
|
|
|
*/
|
|
|
|
module_put(lsock->ops->owner);
|
|
|
|
module_put(sk->sk_prot_creator->owner);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
err:
|
|
|
|
sock_release(lsock);
|
|
|
|
return -EINVAL;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
bool tipc_topsrv_kern_subscr(struct net *net, u32 port, u32 type, u32 lower,
|
|
|
|
u32 upper, u32 filter, int *conid)
|
|
|
|
{
|
|
|
|
struct tipc_subscr sub;
|
|
|
|
struct tipc_conn *con;
|
|
|
|
int rc;
|
|
|
|
|
|
|
|
sub.seq.type = type;
|
|
|
|
sub.seq.lower = lower;
|
|
|
|
sub.seq.upper = upper;
|
|
|
|
sub.timeout = TIPC_WAIT_FOREVER;
|
|
|
|
sub.filter = filter;
|
|
|
|
*(u32 *)&sub.usr_handle = port;
|
|
|
|
|
|
|
|
con = tipc_conn_alloc(tipc_topsrv(net));
|
|
|
|
if (IS_ERR(con))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
*conid = con->conid;
|
|
|
|
con->sock = NULL;
|
|
|
|
rc = tipc_conn_rcv_sub(tipc_topsrv(net), con, &sub);
|
2018-02-19 11:48:21 +00:00
|
|
|
if (rc >= 0)
|
|
|
|
return true;
|
|
|
|
conn_put(con);
|
|
|
|
return false;
|
2018-02-15 09:40:51 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void tipc_topsrv_kern_unsubscr(struct net *net, int conid)
|
|
|
|
{
|
|
|
|
struct tipc_conn *con;
|
|
|
|
|
|
|
|
con = tipc_conn_lookup(tipc_topsrv(net), conid);
|
|
|
|
if (!con)
|
|
|
|
return;
|
|
|
|
|
|
|
|
test_and_clear_bit(CF_CONNECTED, &con->flags);
|
|
|
|
tipc_conn_delete_sub(con, NULL);
|
|
|
|
conn_put(con);
|
|
|
|
conn_put(con);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void tipc_topsrv_kern_evt(struct net *net, struct tipc_event *evt)
|
|
|
|
{
|
|
|
|
u32 port = *(u32 *)&evt->s.usr_handle;
|
|
|
|
u32 self = tipc_own_addr(net);
|
|
|
|
struct sk_buff_head evtq;
|
|
|
|
struct sk_buff *skb;
|
|
|
|
|
|
|
|
skb = tipc_msg_create(TOP_SRV, 0, INT_H_SIZE, sizeof(*evt),
|
|
|
|
self, self, port, port, 0);
|
|
|
|
if (!skb)
|
|
|
|
return;
|
|
|
|
msg_set_dest_droppable(buf_msg(skb), true);
|
|
|
|
memcpy(msg_data(buf_msg(skb)), evt, sizeof(*evt));
|
|
|
|
skb_queue_head_init(&evtq);
|
|
|
|
__skb_queue_tail(&evtq, skb);
|
|
|
|
tipc_sk_rcv(net, &evtq);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int tipc_topsrv_work_start(struct tipc_topsrv *s)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2016-02-02 09:52:17 +00:00
|
|
|
s->rcv_wq = alloc_ordered_workqueue("tipc_rcv", 0);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
if (!s->rcv_wq) {
|
|
|
|
pr_err("can't start tipc receive workqueue\n");
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
2016-02-02 09:52:17 +00:00
|
|
|
s->send_wq = alloc_ordered_workqueue("tipc_send", 0);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
if (!s->send_wq) {
|
|
|
|
pr_err("can't start tipc send workqueue\n");
|
|
|
|
destroy_workqueue(s->rcv_wq);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
static void tipc_topsrv_work_stop(struct tipc_topsrv *s)
|
2018-02-15 09:40:50 +00:00
|
|
|
{
|
|
|
|
destroy_workqueue(s->rcv_wq);
|
|
|
|
destroy_workqueue(s->send_wq);
|
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:49 +00:00
|
|
|
int tipc_topsrv_start(struct net *net)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2018-02-15 09:40:49 +00:00
|
|
|
struct tipc_net *tn = tipc_net(net);
|
|
|
|
const char name[] = "topology_server";
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *srv;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
int ret;
|
|
|
|
|
2018-02-15 09:40:49 +00:00
|
|
|
srv = kzalloc(sizeof(*srv), GFP_ATOMIC);
|
2018-02-15 09:40:50 +00:00
|
|
|
if (!srv)
|
2018-02-15 09:40:49 +00:00
|
|
|
return -ENOMEM;
|
2018-02-15 09:40:50 +00:00
|
|
|
|
|
|
|
srv->net = net;
|
|
|
|
srv->max_rcvbuf_size = sizeof(struct tipc_subscr);
|
2018-02-15 09:40:51 +00:00
|
|
|
INIT_WORK(&srv->awork, tipc_topsrv_accept);
|
2018-02-15 09:40:49 +00:00
|
|
|
|
2018-10-19 04:08:22 +00:00
|
|
|
strscpy(srv->name, name, sizeof(srv->name));
|
2018-02-15 09:40:49 +00:00
|
|
|
tn->topsrv = srv;
|
|
|
|
atomic_set(&tn->subscription_count, 0);
|
|
|
|
|
|
|
|
spin_lock_init(&srv->idr_lock);
|
|
|
|
idr_init(&srv->conn_idr);
|
|
|
|
srv->idr_in_use = 0;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
ret = tipc_topsrv_work_start(srv);
|
2018-02-15 09:40:45 +00:00
|
|
|
if (ret < 0)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
return ret;
|
2018-02-15 09:40:45 +00:00
|
|
|
|
2018-02-15 09:40:51 +00:00
|
|
|
ret = tipc_topsrv_create_listener(srv);
|
2018-02-15 09:40:45 +00:00
|
|
|
if (ret < 0)
|
2018-02-15 09:40:51 +00:00
|
|
|
tipc_topsrv_work_stop(srv);
|
2018-02-15 09:40:45 +00:00
|
|
|
|
2013-08-01 12:29:18 +00:00
|
|
|
return ret;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
|
2018-02-15 09:40:49 +00:00
|
|
|
void tipc_topsrv_stop(struct net *net)
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
{
|
2018-02-15 09:40:51 +00:00
|
|
|
struct tipc_topsrv *srv = tipc_topsrv(net);
|
2018-02-15 09:40:50 +00:00
|
|
|
struct socket *lsock = srv->listener;
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
struct tipc_conn *con;
|
|
|
|
int id;
|
|
|
|
|
2018-02-15 09:40:49 +00:00
|
|
|
spin_lock_bh(&srv->idr_lock);
|
|
|
|
for (id = 0; srv->idr_in_use; id++) {
|
|
|
|
con = idr_find(&srv->conn_idr, id);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
if (con) {
|
2018-02-15 09:40:49 +00:00
|
|
|
spin_unlock_bh(&srv->idr_lock);
|
2018-02-15 09:40:51 +00:00
|
|
|
tipc_conn_close(con);
|
2018-02-15 09:40:49 +00:00
|
|
|
spin_lock_bh(&srv->idr_lock);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|
|
|
|
}
|
2018-02-15 09:40:50 +00:00
|
|
|
__module_get(lsock->ops->owner);
|
|
|
|
__module_get(lsock->sk->sk_prot_creator->owner);
|
|
|
|
srv->listener = NULL;
|
2018-02-15 09:40:49 +00:00
|
|
|
spin_unlock_bh(&srv->idr_lock);
|
tipc: don't call sock_release() in atomic context
syzbot reported a scheduling while atomic issue at netns
destruction time:
BUG: sleeping function called from invalid context at net/core/sock.c:2769
in_atomic(): 1, irqs_disabled(): 0, pid: 85, name: kworker/u4:3
5 locks held by kworker/u4:3/85:
#0: ((wq_completion)"%s""netns"){+.+.}, at: [<00000000c9792deb>]
process_one_work+0xaaf/0x1af0 kernel/workqueue.c:2084
#1: (net_cleanup_work){+.+.}, at: [<00000000adc12e2a>]
process_one_work+0xb01/0x1af0 kernel/workqueue.c:2088
#2: (net_sem){++++}, at: [<000000009ccb5669>] cleanup_net+0x23f/0xd20
net/core/net_namespace.c:494
#3: (net_mutex){+.+.}, at: [<00000000a92767d9>] cleanup_net+0xa7d/0xd20
net/core/net_namespace.c:496
#4: (&(&srv->idr_lock)->rlock){+...}, at: [<000000001343e568>]
spin_lock_bh include/linux/spinlock.h:315 [inline]
#4: (&(&srv->idr_lock)->rlock){+...}, at: [<000000001343e568>]
tipc_topsrv_stop+0x231/0x610 net/tipc/topsrv.c:685
CPU: 0 PID: 85 Comm: kworker/u4:3 Not tainted 4.16.0-rc1+ #230
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS
Google 01/01/2011
Workqueue: netns cleanup_net
Call Trace:
__dump_stack lib/dump_stack.c:17 [inline]
dump_stack+0x194/0x257 lib/dump_stack.c:53
___might_sleep+0x2b2/0x470 kernel/sched/core.c:6128
__might_sleep+0x95/0x190 kernel/sched/core.c:6081
lock_sock_nested+0x37/0x110 net/core/sock.c:2769
lock_sock include/net/sock.h:1463 [inline]
tipc_release+0x103/0xff0 net/tipc/socket.c:572
sock_release+0x8d/0x1e0 net/socket.c:594
tipc_topsrv_stop+0x3c0/0x610 net/tipc/topsrv.c:696
tipc_exit_net+0x15/0x40 net/tipc/core.c:96
ops_exit_list.isra.6+0xae/0x150 net/core/net_namespace.c:148
cleanup_net+0x6ba/0xd20 net/core/net_namespace.c:529
process_one_work+0xbbf/0x1af0 kernel/workqueue.c:2113
worker_thread+0x223/0x1990 kernel/workqueue.c:2247
kthread+0x33c/0x400 kernel/kthread.c:238
ret_from_fork+0x3a/0x50 arch/x86/entry/entry_64.S:429
This is caused by tipc_topsrv_stop() releasing the listener socket
with the idr lock held. This changeset addresses the issue moving
the release operation outside such lock.
Reported-and-tested-by: syzbot+749d9d87c294c00ca856@syzkaller.appspotmail.com
Fixes: 0ef897be12b8 ("tipc: separate topology server listener socket from subcsriber sockets")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Acked-by: ///jon
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-02-19 18:02:24 +00:00
|
|
|
sock_release(lsock);
|
2018-02-15 09:40:51 +00:00
|
|
|
tipc_topsrv_work_stop(srv);
|
2018-02-15 09:40:49 +00:00
|
|
|
idr_destroy(&srv->conn_idr);
|
|
|
|
kfree(srv);
|
tipc: introduce new TIPC server infrastructure
TIPC has two internal servers, one providing a subscription
service for topology events, and another providing the
configuration interface. These servers have previously been running
in BH context, accessing the TIPC-port (aka native) API directly.
Apart from these servers, even the TIPC socket implementation is
partially built on this API.
As this API may simultaneously be called via different paths and in
different contexts, a complex and costly lock policiy is required
in order to protect TIPC internal resources.
To eliminate the need for this complex lock policiy, we introduce
a new, generic service API that uses kernel sockets for message
passing instead of the native API. Once the toplogy and configuration
servers are converted to use this new service, all code pertaining
to the native API can be removed. This entails a significant
reduction in code amount and complexity, and opens up for a complete
rework of the locking policy in TIPC.
The new service also solves another problem:
As the current topology server works in BH context, it cannot easily
be blocked when sending of events fails due to congestion. In such
cases events may have to be silently dropped, something that is
unacceptable. Therefore, the new service keeps a dedicated outbound
queue receiving messages from BH context. Once messages are
inserted into this queue, we will immediately schedule a work from a
special workqueue. This way, messages/events from the topology server
are in reality sent in process context, and the server can block
if necessary.
Analogously, there is a new workqueue for receiving messages. Once a
notification about an arriving message is received in BH context, we
schedule a work from the receive workqueue to do the job of
receiving the message in process context.
As both sending and receive messages are now finished in processes,
subscribed events cannot be dropped any more.
As of this commit, this new server infrastructure is built, but
not actually yet called by the existing TIPC code, but since the
conversion changes required in order to use it are significant,
the addition is kept here as a separate commit.
Signed-off-by: Ying Xue <ying.xue@windriver.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-17 14:54:39 +00:00
|
|
|
}
|