linux_dsm_epyc7002/net/tipc/bcast.c

853 lines
22 KiB
C
Raw Normal View History

/*
* net/tipc/bcast.c: TIPC broadcast code
*
* Copyright (c) 2004-2006, 2014-2017, Ericsson AB
* Copyright (c) 2004, Intel Corporation.
* Copyright (c) 2005, 2010-2011, Wind River Systems
* 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.
*/
#include <linux/tipc_config.h>
#include "socket.h"
#include "msg.h"
#include "bcast.h"
#include "link.h"
#include "name_table.h"
#define BCLINK_WIN_DEFAULT 50 /* bcast link window size (default) */
#define BCLINK_WIN_MIN 32 /* bcast minimum link window size */
const char tipc_bclink_name[] = "broadcast-link";
/**
* struct tipc_bc_base - base structure for keeping broadcast send state
* @link: broadcast send link structure
* @inputq: data input queue; will only carry SOCK_WAKEUP messages
* @dests: array keeping number of reachable destinations per bearer
* @primary_bearer: a bearer having links to all broadcast destinations, if any
* @bcast_support: indicates if primary bearer, if any, supports broadcast
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
* @force_bcast: forces broadcast for multicast traffic
* @rcast_support: indicates if all peer nodes support replicast
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
* @force_rcast: forces replicast for multicast traffic
* @rc_ratio: dest count as percentage of cluster size where send method changes
* @bc_threshold: calculated from rc_ratio; if dests > threshold use broadcast
*/
struct tipc_bc_base {
struct tipc_link *link;
struct sk_buff_head inputq;
int dests[MAX_BEARERS];
int primary_bearer;
bool bcast_support;
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
bool force_bcast;
bool rcast_support;
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
bool force_rcast;
int rc_ratio;
int bc_threshold;
};
static struct tipc_bc_base *tipc_bc_base(struct net *net)
{
return tipc_net(net)->bcbase;
}
/* tipc_bcast_get_mtu(): -get the MTU currently used by broadcast link
* Note: the MTU is decremented to give room for a tunnel header, in
* case the message needs to be sent as replicast
*/
int tipc_bcast_get_mtu(struct net *net)
{
tipc: introduce TIPC encryption & authentication This commit offers an option to encrypt and authenticate all messaging, including the neighbor discovery messages. The currently most advanced algorithm supported is the AEAD AES-GCM (like IPSec or TLS). All encryption/decryption is done at the bearer layer, just before leaving or after entering TIPC. Supported features: - Encryption & authentication of all TIPC messages (header + data); - Two symmetric-key modes: Cluster and Per-node; - Automatic key switching; - Key-expired revoking (sequence number wrapped); - Lock-free encryption/decryption (RCU); - Asynchronous crypto, Intel AES-NI supported; - Multiple cipher transforms; - Logs & statistics; Two key modes: - Cluster key mode: One single key is used for both TX & RX in all nodes in the cluster. - Per-node key mode: Each nodes in the cluster has one specific TX key. For RX, a node requires its peers' TX key to be able to decrypt the messages from those peers. Key setting from user-space is performed via netlink by a user program (e.g. the iproute2 'tipc' tool). Internal key state machine: Attach Align(RX) +-+ +-+ | V | V +---------+ Attach +---------+ | IDLE |---------------->| PENDING |(user = 0) +---------+ +---------+ A A Switch| A | | | | | | Free(switch/revoked) | | (Free)| +----------------------+ | |Timeout | (TX) | | |(RX) | | | | | | v | +---------+ Switch +---------+ | PASSIVE |<----------------| ACTIVE | +---------+ (RX) +---------+ (user = 1) (user >= 1) The number of TFMs is 10 by default and can be changed via the procfs 'net/tipc/max_tfms'. At this moment, as for simplicity, this file is also used to print the crypto statistics at runtime: echo 0xfff1 > /proc/sys/net/tipc/max_tfms The patch defines a new TIPC version (v7) for the encryption message (- backward compatibility as well). The message is basically encapsulated as follows: +----------------------------------------------------------+ | TIPCv7 encryption | Original TIPCv2 | Authentication | | header | packet (encrypted) | Tag | +----------------------------------------------------------+ The throughput is about ~40% for small messages (compared with non- encryption) and ~9% for large messages. With the support from hardware crypto i.e. the Intel AES-NI CPU instructions, the throughput increases upto ~85% for small messages and ~55% for large messages. By default, the new feature is inactive (i.e. no encryption) until user sets a key for TIPC. There is however also a new option - "TIPC_CRYPTO" in the kernel configuration to enable/disable the new code when needed. MAINTAINERS | add two new files 'crypto.h' & 'crypto.c' in tipc Acked-by: Ying Xue <ying.xue@windreiver.com> Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-11-08 12:05:11 +07:00
return tipc_link_mss(tipc_bc_sndlink(net));
}
void tipc_bcast_toggle_rcast(struct net *net, bool supp)
{
tipc_bc_base(net)->rcast_support = supp;
}
static void tipc_bcbase_calc_bc_threshold(struct net *net)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
int cluster_size = tipc_link_bc_peers(tipc_bc_sndlink(net));
bb->bc_threshold = 1 + (cluster_size * bb->rc_ratio / 100);
}
/* tipc_bcbase_select_primary(): find a bearer with links to all destinations,
* if any, and make it primary bearer
*/
static void tipc_bcbase_select_primary(struct net *net)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
int all_dests = tipc_link_bc_peers(bb->link);
int i, mtu, prim;
bb->primary_bearer = INVALID_BEARER_ID;
bb->bcast_support = true;
if (!all_dests)
return;
for (i = 0; i < MAX_BEARERS; i++) {
if (!bb->dests[i])
continue;
mtu = tipc_bearer_mtu(net, i);
if (mtu < tipc_link_mtu(bb->link))
tipc_link_set_mtu(bb->link, mtu);
bb->bcast_support &= tipc_bearer_bcast_support(net, i);
if (bb->dests[i] < all_dests)
continue;
bb->primary_bearer = i;
/* Reduce risk that all nodes select same primary */
if ((i ^ tipc_own_addr(net)) & 1)
break;
}
prim = bb->primary_bearer;
if (prim != INVALID_BEARER_ID)
bb->bcast_support = tipc_bearer_bcast_support(net, prim);
}
void tipc_bcast_inc_bearer_dst_cnt(struct net *net, int bearer_id)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
tipc_bcast_lock(net);
bb->dests[bearer_id]++;
tipc_bcbase_select_primary(net);
tipc_bcast_unlock(net);
}
void tipc_bcast_dec_bearer_dst_cnt(struct net *net, int bearer_id)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
tipc_bcast_lock(net);
bb->dests[bearer_id]--;
tipc_bcbase_select_primary(net);
tipc_bcast_unlock(net);
}
/* tipc_bcbase_xmit - broadcast a packet queue across one or more bearers
*
* Note that number of reachable destinations, as indicated in the dests[]
* array, may transitionally differ from the number of destinations indicated
* in each sent buffer. We can sustain this. Excess destination nodes will
* drop and never acknowledge the unexpected packets, and missing destinations
* will either require retransmission (if they are just about to be added to
* the bearer), or be removed from the buffer's 'ackers' counter (if they
* just went down)
*/
static void tipc_bcbase_xmit(struct net *net, struct sk_buff_head *xmitq)
{
int bearer_id;
struct tipc_bc_base *bb = tipc_bc_base(net);
struct sk_buff *skb, *_skb;
struct sk_buff_head _xmitq;
if (skb_queue_empty(xmitq))
return;
/* The typical case: at least one bearer has links to all nodes */
bearer_id = bb->primary_bearer;
if (bearer_id >= 0) {
tipc_bearer_bc_xmit(net, bearer_id, xmitq);
return;
}
/* We have to transmit across all bearers */
tipc: clean up skb list lock handling on send path The policy for handling the skb list locks on the send and receive paths is simple. - On the send path we never need to grab the lock on the 'xmitq' list when the destination is an exernal node. - On the receive path we always need to grab the lock on the 'inputq' list, irrespective of source node. However, when transmitting node local messages those will eventually end up on the receive path of a local socket, meaning that the argument 'xmitq' in tipc_node_xmit() will become the 'ínputq' argument in the function tipc_sk_rcv(). This has been handled by always initializing the spinlock of the 'xmitq' list at message creation, just in case it may end up on the receive path later, and despite knowing that the lock in most cases never will be used. This approach is inaccurate and confusing, and has also concealed the fact that the stated 'no lock grabbing' policy for the send path is violated in some cases. We now clean up this by never initializing the lock at message creation, instead doing this at the moment we find that the message actually will enter the receive path. At the same time we fix the four locations where we incorrectly access the spinlock on the send/error path. This patch also reverts commit d12cffe9329f ("tipc: ensure head->lock is initialised") which has now become redundant. CC: Eric Dumazet <edumazet@google.com> Reported-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Reviewed-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-08-15 21:42:50 +07:00
__skb_queue_head_init(&_xmitq);
for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
if (!bb->dests[bearer_id])
continue;
skb_queue_walk(xmitq, skb) {
_skb = pskb_copy_for_clone(skb, GFP_ATOMIC);
if (!_skb)
break;
__skb_queue_tail(&_xmitq, _skb);
}
tipc_bearer_bc_xmit(net, bearer_id, &_xmitq);
}
__skb_queue_purge(xmitq);
__skb_queue_purge(&_xmitq);
}
static void tipc_bcast_select_xmit_method(struct net *net, int dests,
struct tipc_mc_method *method)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
unsigned long exp = method->expires;
/* Broadcast supported by used bearer/bearers? */
if (!bb->bcast_support) {
method->rcast = true;
return;
}
/* Any destinations which don't support replicast ? */
if (!bb->rcast_support) {
method->rcast = false;
return;
}
/* Can current method be changed ? */
method->expires = jiffies + TIPC_METHOD_EXPIRE;
if (method->mandatory)
return;
if (!(tipc_net(net)->capabilities & TIPC_MCAST_RBCTL) &&
time_before(jiffies, exp))
return;
/* Configuration as force 'broadcast' method */
if (bb->force_bcast) {
method->rcast = false;
return;
}
/* Configuration as force 'replicast' method */
if (bb->force_rcast) {
method->rcast = true;
return;
}
/* Configuration as 'autoselect' or default method */
/* Determine method to use now */
method->rcast = dests <= bb->bc_threshold;
}
/* tipc_bcast_xmit - broadcast the buffer chain to all external nodes
* @net: the applicable net namespace
* @pkts: chain of buffers containing message
* @cong_link_cnt: set to 1 if broadcast link is congested, otherwise 0
tipc: reduce risk of user starvation during link congestion The socket code currently handles link congestion by either blocking and trying to send again when the congestion has abated, or just returning to the user with -EAGAIN and let him re-try later. This mechanism is prone to starvation, because the wakeup algorithm is non-atomic. During the time the link issues a wakeup signal, until the socket wakes up and re-attempts sending, other senders may have come in between and occupied the free buffer space in the link. This in turn may lead to a socket having to make many send attempts before it is successful. In extremely loaded systems we have observed latency times of several seconds before a low-priority socket is able to send out a message. In this commit, we simplify this mechanism and reduce the risk of the described scenario happening. When a message is attempted sent via a congested link, we now let it be added to the link's backlog queue anyway, thus permitting an oversubscription of one message per source socket. We still create a wakeup item and return an error code, hence instructing the sender to block or stop sending. Only when enough space has been freed up in the link's backlog queue do we issue a wakeup event that allows the sender to continue with the next message, if any. The fact that a socket now can consider a message sent even when the link returns a congestion code means that the sending socket code can be simplified. Also, since this is a good opportunity to get rid of the obsolete 'mtu change' condition in the three socket send functions, we now choose to refactor those functions completely. Signed-off-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@ericsson.com> 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>
2017-01-03 22:55:11 +07:00
* Consumes the buffer chain.
* Returns 0 if success, otherwise errno: -EHOSTUNREACH,-EMSGSIZE
*/
static int tipc_bcast_xmit(struct net *net, struct sk_buff_head *pkts,
u16 *cong_link_cnt)
{
struct tipc_link *l = tipc_bc_sndlink(net);
struct sk_buff_head xmitq;
int rc = 0;
tipc: clean up skb list lock handling on send path The policy for handling the skb list locks on the send and receive paths is simple. - On the send path we never need to grab the lock on the 'xmitq' list when the destination is an exernal node. - On the receive path we always need to grab the lock on the 'inputq' list, irrespective of source node. However, when transmitting node local messages those will eventually end up on the receive path of a local socket, meaning that the argument 'xmitq' in tipc_node_xmit() will become the 'ínputq' argument in the function tipc_sk_rcv(). This has been handled by always initializing the spinlock of the 'xmitq' list at message creation, just in case it may end up on the receive path later, and despite knowing that the lock in most cases never will be used. This approach is inaccurate and confusing, and has also concealed the fact that the stated 'no lock grabbing' policy for the send path is violated in some cases. We now clean up this by never initializing the lock at message creation, instead doing this at the moment we find that the message actually will enter the receive path. At the same time we fix the four locations where we incorrectly access the spinlock on the send/error path. This patch also reverts commit d12cffe9329f ("tipc: ensure head->lock is initialised") which has now become redundant. CC: Eric Dumazet <edumazet@google.com> Reported-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Reviewed-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-08-15 21:42:50 +07:00
__skb_queue_head_init(&xmitq);
tipc_bcast_lock(net);
if (tipc_link_bc_peers(l))
rc = tipc_link_xmit(l, pkts, &xmitq);
tipc_bcast_unlock(net);
tipc_bcbase_xmit(net, &xmitq);
__skb_queue_purge(pkts);
if (rc == -ELINKCONG) {
*cong_link_cnt = 1;
rc = 0;
}
return rc;
}
/* tipc_rcast_xmit - replicate and send a message to given destination nodes
* @net: the applicable net namespace
* @pkts: chain of buffers containing message
* @dests: list of destination nodes
* @cong_link_cnt: returns number of congested links
* @cong_links: returns identities of congested links
* Returns 0 if success, otherwise errno
*/
static int tipc_rcast_xmit(struct net *net, struct sk_buff_head *pkts,
struct tipc_nlist *dests, u16 *cong_link_cnt)
{
struct tipc_dest *dst, *tmp;
struct sk_buff_head _pkts;
u32 dnode, selector;
selector = msg_link_selector(buf_msg(skb_peek(pkts)));
tipc: clean up skb list lock handling on send path The policy for handling the skb list locks on the send and receive paths is simple. - On the send path we never need to grab the lock on the 'xmitq' list when the destination is an exernal node. - On the receive path we always need to grab the lock on the 'inputq' list, irrespective of source node. However, when transmitting node local messages those will eventually end up on the receive path of a local socket, meaning that the argument 'xmitq' in tipc_node_xmit() will become the 'ínputq' argument in the function tipc_sk_rcv(). This has been handled by always initializing the spinlock of the 'xmitq' list at message creation, just in case it may end up on the receive path later, and despite knowing that the lock in most cases never will be used. This approach is inaccurate and confusing, and has also concealed the fact that the stated 'no lock grabbing' policy for the send path is violated in some cases. We now clean up this by never initializing the lock at message creation, instead doing this at the moment we find that the message actually will enter the receive path. At the same time we fix the four locations where we incorrectly access the spinlock on the send/error path. This patch also reverts commit d12cffe9329f ("tipc: ensure head->lock is initialised") which has now become redundant. CC: Eric Dumazet <edumazet@google.com> Reported-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Reviewed-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-08-15 21:42:50 +07:00
__skb_queue_head_init(&_pkts);
list_for_each_entry_safe(dst, tmp, &dests->list, list) {
dnode = dst->node;
if (!tipc_msg_pskb_copy(dnode, pkts, &_pkts))
return -ENOMEM;
/* Any other return value than -ELINKCONG is ignored */
if (tipc_node_xmit(net, &_pkts, dnode, selector) == -ELINKCONG)
(*cong_link_cnt)++;
}
return 0;
}
/* tipc_mcast_send_sync - deliver a dummy message with SYN bit
* @net: the applicable net namespace
* @skb: socket buffer to copy
* @method: send method to be used
* @dests: destination nodes for message.
* Returns 0 if success, otherwise errno
*/
static int tipc_mcast_send_sync(struct net *net, struct sk_buff *skb,
struct tipc_mc_method *method,
tipc: fix potential hanging after b/rcast changing In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast"), we allow instant switching between replicast and broadcast by sending a dummy 'SYN' packet on the last used link to synchronize packets on the links. The 'SYN' message is an object of link congestion also, so if that happens, a 'SOCK_WAKEUP' will be scheduled to be sent back to the socket... However, in that commit, we simply use the same socket 'cong_link_cnt' counter for both the 'SYN' & normal payload message sending. Therefore, if both the replicast & broadcast links are congested, the counter will be not updated correctly but overwritten by the latter congestion. Later on, when the 'SOCK_WAKEUP' messages are processed, the counter is reduced one by one and eventually overflowed. Consequently, further activities on the socket will only wait for the false congestion signal to disappear but never been met. Because sending the 'SYN' message is vital for the mechanism, it should be done anyway. This commit fixes the issue by marking the message with an error code e.g. 'TIPC_ERR_NO_PORT', so its sending should not face a link congestion, there is no need to touch the socket 'cong_link_cnt' either. In addition, in the event of any error (e.g. -ENOBUFS), we will purge the entire payload message queue and make a return immediately. Fixes: c55c8edafa91 ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-10 15:21:03 +07:00
struct tipc_nlist *dests)
{
struct tipc_msg *hdr, *_hdr;
struct sk_buff_head tmpq;
struct sk_buff *_skb;
tipc: fix potential hanging after b/rcast changing In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast"), we allow instant switching between replicast and broadcast by sending a dummy 'SYN' packet on the last used link to synchronize packets on the links. The 'SYN' message is an object of link congestion also, so if that happens, a 'SOCK_WAKEUP' will be scheduled to be sent back to the socket... However, in that commit, we simply use the same socket 'cong_link_cnt' counter for both the 'SYN' & normal payload message sending. Therefore, if both the replicast & broadcast links are congested, the counter will be not updated correctly but overwritten by the latter congestion. Later on, when the 'SOCK_WAKEUP' messages are processed, the counter is reduced one by one and eventually overflowed. Consequently, further activities on the socket will only wait for the false congestion signal to disappear but never been met. Because sending the 'SYN' message is vital for the mechanism, it should be done anyway. This commit fixes the issue by marking the message with an error code e.g. 'TIPC_ERR_NO_PORT', so its sending should not face a link congestion, there is no need to touch the socket 'cong_link_cnt' either. In addition, in the event of any error (e.g. -ENOBUFS), we will purge the entire payload message queue and make a return immediately. Fixes: c55c8edafa91 ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-10 15:21:03 +07:00
u16 cong_link_cnt;
int rc = 0;
/* Is a cluster supporting with new capabilities ? */
if (!(tipc_net(net)->capabilities & TIPC_MCAST_RBCTL))
return 0;
hdr = buf_msg(skb);
if (msg_user(hdr) == MSG_FRAGMENTER)
hdr = msg_inner_hdr(hdr);
if (msg_type(hdr) != TIPC_MCAST_MSG)
return 0;
/* Allocate dummy message */
_skb = tipc_buf_acquire(MCAST_H_SIZE, GFP_KERNEL);
if (!_skb)
return -ENOMEM;
/* Preparing for 'synching' header */
msg_set_syn(hdr, 1);
/* Copy skb's header into a dummy header */
skb_copy_to_linear_data(_skb, hdr, MCAST_H_SIZE);
skb_orphan(_skb);
/* Reverse method for dummy message */
_hdr = buf_msg(_skb);
msg_set_size(_hdr, MCAST_H_SIZE);
msg_set_is_rcast(_hdr, !msg_is_rcast(hdr));
tipc: fix potential hanging after b/rcast changing In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast"), we allow instant switching between replicast and broadcast by sending a dummy 'SYN' packet on the last used link to synchronize packets on the links. The 'SYN' message is an object of link congestion also, so if that happens, a 'SOCK_WAKEUP' will be scheduled to be sent back to the socket... However, in that commit, we simply use the same socket 'cong_link_cnt' counter for both the 'SYN' & normal payload message sending. Therefore, if both the replicast & broadcast links are congested, the counter will be not updated correctly but overwritten by the latter congestion. Later on, when the 'SOCK_WAKEUP' messages are processed, the counter is reduced one by one and eventually overflowed. Consequently, further activities on the socket will only wait for the false congestion signal to disappear but never been met. Because sending the 'SYN' message is vital for the mechanism, it should be done anyway. This commit fixes the issue by marking the message with an error code e.g. 'TIPC_ERR_NO_PORT', so its sending should not face a link congestion, there is no need to touch the socket 'cong_link_cnt' either. In addition, in the event of any error (e.g. -ENOBUFS), we will purge the entire payload message queue and make a return immediately. Fixes: c55c8edafa91 ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-10 15:21:03 +07:00
msg_set_errcode(_hdr, TIPC_ERR_NO_PORT);
tipc: clean up skb list lock handling on send path The policy for handling the skb list locks on the send and receive paths is simple. - On the send path we never need to grab the lock on the 'xmitq' list when the destination is an exernal node. - On the receive path we always need to grab the lock on the 'inputq' list, irrespective of source node. However, when transmitting node local messages those will eventually end up on the receive path of a local socket, meaning that the argument 'xmitq' in tipc_node_xmit() will become the 'ínputq' argument in the function tipc_sk_rcv(). This has been handled by always initializing the spinlock of the 'xmitq' list at message creation, just in case it may end up on the receive path later, and despite knowing that the lock in most cases never will be used. This approach is inaccurate and confusing, and has also concealed the fact that the stated 'no lock grabbing' policy for the send path is violated in some cases. We now clean up this by never initializing the lock at message creation, instead doing this at the moment we find that the message actually will enter the receive path. At the same time we fix the four locations where we incorrectly access the spinlock on the send/error path. This patch also reverts commit d12cffe9329f ("tipc: ensure head->lock is initialised") which has now become redundant. CC: Eric Dumazet <edumazet@google.com> Reported-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Reviewed-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-08-15 21:42:50 +07:00
__skb_queue_head_init(&tmpq);
__skb_queue_tail(&tmpq, _skb);
if (method->rcast)
tipc: fix potential hanging after b/rcast changing In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast"), we allow instant switching between replicast and broadcast by sending a dummy 'SYN' packet on the last used link to synchronize packets on the links. The 'SYN' message is an object of link congestion also, so if that happens, a 'SOCK_WAKEUP' will be scheduled to be sent back to the socket... However, in that commit, we simply use the same socket 'cong_link_cnt' counter for both the 'SYN' & normal payload message sending. Therefore, if both the replicast & broadcast links are congested, the counter will be not updated correctly but overwritten by the latter congestion. Later on, when the 'SOCK_WAKEUP' messages are processed, the counter is reduced one by one and eventually overflowed. Consequently, further activities on the socket will only wait for the false congestion signal to disappear but never been met. Because sending the 'SYN' message is vital for the mechanism, it should be done anyway. This commit fixes the issue by marking the message with an error code e.g. 'TIPC_ERR_NO_PORT', so its sending should not face a link congestion, there is no need to touch the socket 'cong_link_cnt' either. In addition, in the event of any error (e.g. -ENOBUFS), we will purge the entire payload message queue and make a return immediately. Fixes: c55c8edafa91 ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-10 15:21:03 +07:00
rc = tipc_bcast_xmit(net, &tmpq, &cong_link_cnt);
else
tipc: fix potential hanging after b/rcast changing In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast"), we allow instant switching between replicast and broadcast by sending a dummy 'SYN' packet on the last used link to synchronize packets on the links. The 'SYN' message is an object of link congestion also, so if that happens, a 'SOCK_WAKEUP' will be scheduled to be sent back to the socket... However, in that commit, we simply use the same socket 'cong_link_cnt' counter for both the 'SYN' & normal payload message sending. Therefore, if both the replicast & broadcast links are congested, the counter will be not updated correctly but overwritten by the latter congestion. Later on, when the 'SOCK_WAKEUP' messages are processed, the counter is reduced one by one and eventually overflowed. Consequently, further activities on the socket will only wait for the false congestion signal to disappear but never been met. Because sending the 'SYN' message is vital for the mechanism, it should be done anyway. This commit fixes the issue by marking the message with an error code e.g. 'TIPC_ERR_NO_PORT', so its sending should not face a link congestion, there is no need to touch the socket 'cong_link_cnt' either. In addition, in the event of any error (e.g. -ENOBUFS), we will purge the entire payload message queue and make a return immediately. Fixes: c55c8edafa91 ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-10 15:21:03 +07:00
rc = tipc_rcast_xmit(net, &tmpq, dests, &cong_link_cnt);
/* This queue should normally be empty by now */
__skb_queue_purge(&tmpq);
tipc: fix potential hanging after b/rcast changing In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast"), we allow instant switching between replicast and broadcast by sending a dummy 'SYN' packet on the last used link to synchronize packets on the links. The 'SYN' message is an object of link congestion also, so if that happens, a 'SOCK_WAKEUP' will be scheduled to be sent back to the socket... However, in that commit, we simply use the same socket 'cong_link_cnt' counter for both the 'SYN' & normal payload message sending. Therefore, if both the replicast & broadcast links are congested, the counter will be not updated correctly but overwritten by the latter congestion. Later on, when the 'SOCK_WAKEUP' messages are processed, the counter is reduced one by one and eventually overflowed. Consequently, further activities on the socket will only wait for the false congestion signal to disappear but never been met. Because sending the 'SYN' message is vital for the mechanism, it should be done anyway. This commit fixes the issue by marking the message with an error code e.g. 'TIPC_ERR_NO_PORT', so its sending should not face a link congestion, there is no need to touch the socket 'cong_link_cnt' either. In addition, in the event of any error (e.g. -ENOBUFS), we will purge the entire payload message queue and make a return immediately. Fixes: c55c8edafa91 ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-10 15:21:03 +07:00
return rc;
}
/* tipc_mcast_xmit - deliver message to indicated destination nodes
* and to identified node local sockets
* @net: the applicable net namespace
* @pkts: chain of buffers containing message
* @method: send method to be used
* @dests: destination nodes for message.
* @cong_link_cnt: returns number of encountered congested destination links
* Consumes buffer chain.
* Returns 0 if success, otherwise errno
*/
int tipc_mcast_xmit(struct net *net, struct sk_buff_head *pkts,
struct tipc_mc_method *method, struct tipc_nlist *dests,
u16 *cong_link_cnt)
{
struct sk_buff_head inputq, localq;
bool rcast = method->rcast;
struct tipc_msg *hdr;
struct sk_buff *skb;
int rc = 0;
skb_queue_head_init(&inputq);
tipc: clean up skb list lock handling on send path The policy for handling the skb list locks on the send and receive paths is simple. - On the send path we never need to grab the lock on the 'xmitq' list when the destination is an exernal node. - On the receive path we always need to grab the lock on the 'inputq' list, irrespective of source node. However, when transmitting node local messages those will eventually end up on the receive path of a local socket, meaning that the argument 'xmitq' in tipc_node_xmit() will become the 'ínputq' argument in the function tipc_sk_rcv(). This has been handled by always initializing the spinlock of the 'xmitq' list at message creation, just in case it may end up on the receive path later, and despite knowing that the lock in most cases never will be used. This approach is inaccurate and confusing, and has also concealed the fact that the stated 'no lock grabbing' policy for the send path is violated in some cases. We now clean up this by never initializing the lock at message creation, instead doing this at the moment we find that the message actually will enter the receive path. At the same time we fix the four locations where we incorrectly access the spinlock on the send/error path. This patch also reverts commit d12cffe9329f ("tipc: ensure head->lock is initialised") which has now become redundant. CC: Eric Dumazet <edumazet@google.com> Reported-by: Chris Packham <chris.packham@alliedtelesis.co.nz> Acked-by: Ying Xue <ying.xue@windriver.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Reviewed-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-08-15 21:42:50 +07:00
__skb_queue_head_init(&localq);
/* Clone packets before they are consumed by next call */
if (dests->local && !tipc_msg_reassemble(pkts, &localq)) {
rc = -ENOMEM;
goto exit;
}
/* Send according to determined transmit method */
if (dests->remote) {
tipc_bcast_select_xmit_method(net, dests->remote, method);
skb = skb_peek(pkts);
hdr = buf_msg(skb);
if (msg_user(hdr) == MSG_FRAGMENTER)
hdr = msg_inner_hdr(hdr);
msg_set_is_rcast(hdr, method->rcast);
/* Switch method ? */
tipc: fix potential hanging after b/rcast changing In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast"), we allow instant switching between replicast and broadcast by sending a dummy 'SYN' packet on the last used link to synchronize packets on the links. The 'SYN' message is an object of link congestion also, so if that happens, a 'SOCK_WAKEUP' will be scheduled to be sent back to the socket... However, in that commit, we simply use the same socket 'cong_link_cnt' counter for both the 'SYN' & normal payload message sending. Therefore, if both the replicast & broadcast links are congested, the counter will be not updated correctly but overwritten by the latter congestion. Later on, when the 'SOCK_WAKEUP' messages are processed, the counter is reduced one by one and eventually overflowed. Consequently, further activities on the socket will only wait for the false congestion signal to disappear but never been met. Because sending the 'SYN' message is vital for the mechanism, it should be done anyway. This commit fixes the issue by marking the message with an error code e.g. 'TIPC_ERR_NO_PORT', so its sending should not face a link congestion, there is no need to touch the socket 'cong_link_cnt' either. In addition, in the event of any error (e.g. -ENOBUFS), we will purge the entire payload message queue and make a return immediately. Fixes: c55c8edafa91 ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Tuong Lien <tuong.t.lien@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-10 15:21:03 +07:00
if (rcast != method->rcast) {
rc = tipc_mcast_send_sync(net, skb, method, dests);
if (unlikely(rc)) {
pr_err("Unable to send SYN: method %d, rc %d\n",
rcast, rc);
goto exit;
}
}
if (method->rcast)
rc = tipc_rcast_xmit(net, pkts, dests, cong_link_cnt);
else
rc = tipc_bcast_xmit(net, pkts, cong_link_cnt);
}
if (dests->local) {
tipc_loopback_trace(net, &localq);
tipc_sk_mcast_rcv(net, &localq, &inputq);
}
exit:
/* This queue should normally be empty by now */
__skb_queue_purge(pkts);
tipc: reduce risk of user starvation during link congestion The socket code currently handles link congestion by either blocking and trying to send again when the congestion has abated, or just returning to the user with -EAGAIN and let him re-try later. This mechanism is prone to starvation, because the wakeup algorithm is non-atomic. During the time the link issues a wakeup signal, until the socket wakes up and re-attempts sending, other senders may have come in between and occupied the free buffer space in the link. This in turn may lead to a socket having to make many send attempts before it is successful. In extremely loaded systems we have observed latency times of several seconds before a low-priority socket is able to send out a message. In this commit, we simplify this mechanism and reduce the risk of the described scenario happening. When a message is attempted sent via a congested link, we now let it be added to the link's backlog queue anyway, thus permitting an oversubscription of one message per source socket. We still create a wakeup item and return an error code, hence instructing the sender to block or stop sending. Only when enough space has been freed up in the link's backlog queue do we issue a wakeup event that allows the sender to continue with the next message, if any. The fact that a socket now can consider a message sent even when the link returns a congestion code means that the sending socket code can be simplified. Also, since this is a good opportunity to get rid of the obsolete 'mtu change' condition in the three socket send functions, we now choose to refactor those functions completely. Signed-off-by: Parthasarathy Bhuvaragan <parthasarathy.bhuvaragan@ericsson.com> 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>
2017-01-03 22:55:11 +07:00
return rc;
}
/* tipc_bcast_rcv - receive a broadcast packet, and deliver to rcv link
*
* RCU is locked, no other locks set
*/
int tipc_bcast_rcv(struct net *net, struct tipc_link *l, struct sk_buff *skb)
{
struct tipc_msg *hdr = buf_msg(skb);
struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
struct sk_buff_head xmitq;
int rc;
__skb_queue_head_init(&xmitq);
if (msg_mc_netid(hdr) != tipc_netid(net) || !tipc_link_is_up(l)) {
kfree_skb(skb);
return 0;
}
tipc_bcast_lock(net);
if (msg_user(hdr) == BCAST_PROTOCOL)
rc = tipc_link_bc_nack_rcv(l, skb, &xmitq);
else
rc = tipc_link_rcv(l, skb, NULL);
tipc_bcast_unlock(net);
tipc_bcbase_xmit(net, &xmitq);
/* Any socket wakeup messages ? */
if (!skb_queue_empty(inputq))
tipc_sk_rcv(net, inputq);
return rc;
}
/* tipc_bcast_ack_rcv - receive and handle a broadcast acknowledge
*
* RCU is locked, no other locks set
*/
tipc: fix broadcast link synchronization problem In commit 2d18ac4ba745 ("tipc: extend broadcast link initialization criteria") we tried to fix a problem with the initial synchronization of broadcast link acknowledge values. Unfortunately that solution is not sufficient to solve the issue. We have seen it happen that LINK_PROTOCOL/STATE packets with a valid non-zero unicast acknowledge number may bypass BCAST_PROTOCOL initialization, NAME_DISTRIBUTOR and other STATE packets with invalid broadcast acknowledge numbers, leading to premature opening of the broadcast link. When the bypassed packets finally arrive, they are inadvertently accepted, and the already correctly initialized acknowledge number in the broadcast receive link is overwritten by the invalid (zero) value of the said packets. After this the broadcast link goes stale. We now fix this by marking the packets where we know the acknowledge value is or may be invalid, and then ignoring the acks from those. To this purpose, we claim an unused bit in the header to indicate that the value is invalid. We set the bit to 1 in the initial BCAST_PROTOCOL synchronization packet and all initial ("bulk") NAME_DISTRIBUTOR packets, plus those LINK_PROTOCOL packets sent out before the broadcast links are fully synchronized. This minor protocol update is fully backwards compatible. Reported-by: John Thompson <thompa.atl@gmail.com> Tested-by: John Thompson <thompa.atl@gmail.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-10-28 05:51:55 +07:00
void tipc_bcast_ack_rcv(struct net *net, struct tipc_link *l,
struct tipc_msg *hdr)
{
struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
tipc: fix broadcast link synchronization problem In commit 2d18ac4ba745 ("tipc: extend broadcast link initialization criteria") we tried to fix a problem with the initial synchronization of broadcast link acknowledge values. Unfortunately that solution is not sufficient to solve the issue. We have seen it happen that LINK_PROTOCOL/STATE packets with a valid non-zero unicast acknowledge number may bypass BCAST_PROTOCOL initialization, NAME_DISTRIBUTOR and other STATE packets with invalid broadcast acknowledge numbers, leading to premature opening of the broadcast link. When the bypassed packets finally arrive, they are inadvertently accepted, and the already correctly initialized acknowledge number in the broadcast receive link is overwritten by the invalid (zero) value of the said packets. After this the broadcast link goes stale. We now fix this by marking the packets where we know the acknowledge value is or may be invalid, and then ignoring the acks from those. To this purpose, we claim an unused bit in the header to indicate that the value is invalid. We set the bit to 1 in the initial BCAST_PROTOCOL synchronization packet and all initial ("bulk") NAME_DISTRIBUTOR packets, plus those LINK_PROTOCOL packets sent out before the broadcast links are fully synchronized. This minor protocol update is fully backwards compatible. Reported-by: John Thompson <thompa.atl@gmail.com> Tested-by: John Thompson <thompa.atl@gmail.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-10-28 05:51:55 +07:00
u16 acked = msg_bcast_ack(hdr);
struct sk_buff_head xmitq;
tipc: fix broadcast link synchronization problem In commit 2d18ac4ba745 ("tipc: extend broadcast link initialization criteria") we tried to fix a problem with the initial synchronization of broadcast link acknowledge values. Unfortunately that solution is not sufficient to solve the issue. We have seen it happen that LINK_PROTOCOL/STATE packets with a valid non-zero unicast acknowledge number may bypass BCAST_PROTOCOL initialization, NAME_DISTRIBUTOR and other STATE packets with invalid broadcast acknowledge numbers, leading to premature opening of the broadcast link. When the bypassed packets finally arrive, they are inadvertently accepted, and the already correctly initialized acknowledge number in the broadcast receive link is overwritten by the invalid (zero) value of the said packets. After this the broadcast link goes stale. We now fix this by marking the packets where we know the acknowledge value is or may be invalid, and then ignoring the acks from those. To this purpose, we claim an unused bit in the header to indicate that the value is invalid. We set the bit to 1 in the initial BCAST_PROTOCOL synchronization packet and all initial ("bulk") NAME_DISTRIBUTOR packets, plus those LINK_PROTOCOL packets sent out before the broadcast links are fully synchronized. This minor protocol update is fully backwards compatible. Reported-by: John Thompson <thompa.atl@gmail.com> Tested-by: John Thompson <thompa.atl@gmail.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-10-28 05:51:55 +07:00
/* Ignore bc acks sent by peer before bcast synch point was received */
if (msg_bc_ack_invalid(hdr))
return;
__skb_queue_head_init(&xmitq);
tipc_bcast_lock(net);
tipc_link_bc_ack_rcv(l, acked, &xmitq);
tipc_bcast_unlock(net);
tipc_bcbase_xmit(net, &xmitq);
/* Any socket wakeup messages ? */
if (!skb_queue_empty(inputq))
tipc_sk_rcv(net, inputq);
}
/* tipc_bcast_synch_rcv - check and update rcv link with peer's send state
*
* RCU is locked, no other locks set
*/
int tipc_bcast_sync_rcv(struct net *net, struct tipc_link *l,
struct tipc_msg *hdr)
{
struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
struct sk_buff_head xmitq;
int rc = 0;
__skb_queue_head_init(&xmitq);
tipc_bcast_lock(net);
tipc: fix broadcast link synchronization problem In commit 2d18ac4ba745 ("tipc: extend broadcast link initialization criteria") we tried to fix a problem with the initial synchronization of broadcast link acknowledge values. Unfortunately that solution is not sufficient to solve the issue. We have seen it happen that LINK_PROTOCOL/STATE packets with a valid non-zero unicast acknowledge number may bypass BCAST_PROTOCOL initialization, NAME_DISTRIBUTOR and other STATE packets with invalid broadcast acknowledge numbers, leading to premature opening of the broadcast link. When the bypassed packets finally arrive, they are inadvertently accepted, and the already correctly initialized acknowledge number in the broadcast receive link is overwritten by the invalid (zero) value of the said packets. After this the broadcast link goes stale. We now fix this by marking the packets where we know the acknowledge value is or may be invalid, and then ignoring the acks from those. To this purpose, we claim an unused bit in the header to indicate that the value is invalid. We set the bit to 1 in the initial BCAST_PROTOCOL synchronization packet and all initial ("bulk") NAME_DISTRIBUTOR packets, plus those LINK_PROTOCOL packets sent out before the broadcast links are fully synchronized. This minor protocol update is fully backwards compatible. Reported-by: John Thompson <thompa.atl@gmail.com> Tested-by: John Thompson <thompa.atl@gmail.com> Signed-off-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2016-10-28 05:51:55 +07:00
if (msg_type(hdr) != STATE_MSG) {
tipc_link_bc_init_rcv(l, hdr);
} else if (!msg_bc_ack_invalid(hdr)) {
tipc_link_bc_ack_rcv(l, msg_bcast_ack(hdr), &xmitq);
rc = tipc_link_bc_sync_rcv(l, hdr, &xmitq);
}
tipc_bcast_unlock(net);
tipc_bcbase_xmit(net, &xmitq);
/* Any socket wakeup messages ? */
if (!skb_queue_empty(inputq))
tipc_sk_rcv(net, inputq);
return rc;
}
/* tipc_bcast_add_peer - add a peer node to broadcast link and bearer
*
* RCU is locked, node lock is set
*/
void tipc_bcast_add_peer(struct net *net, struct tipc_link *uc_l,
struct sk_buff_head *xmitq)
{
struct tipc_link *snd_l = tipc_bc_sndlink(net);
tipc_bcast_lock(net);
tipc_link_add_bc_peer(snd_l, uc_l, xmitq);
tipc_bcbase_select_primary(net);
tipc_bcbase_calc_bc_threshold(net);
tipc_bcast_unlock(net);
}
/* tipc_bcast_remove_peer - remove a peer node from broadcast link and bearer
*
* RCU is locked, node lock is set
*/
void tipc_bcast_remove_peer(struct net *net, struct tipc_link *rcv_l)
{
struct tipc_link *snd_l = tipc_bc_sndlink(net);
struct sk_buff_head *inputq = &tipc_bc_base(net)->inputq;
struct sk_buff_head xmitq;
__skb_queue_head_init(&xmitq);
tipc_bcast_lock(net);
tipc_link_remove_bc_peer(snd_l, rcv_l, &xmitq);
tipc_bcbase_select_primary(net);
tipc_bcbase_calc_bc_threshold(net);
tipc_bcast_unlock(net);
tipc_bcbase_xmit(net, &xmitq);
/* Any socket wakeup messages ? */
if (!skb_queue_empty(inputq))
tipc_sk_rcv(net, inputq);
}
int tipc_bclink_reset_stats(struct net *net)
{
struct tipc_link *l = tipc_bc_sndlink(net);
if (!l)
return -ENOPROTOOPT;
tipc_bcast_lock(net);
tipc_link_reset_stats(l);
tipc_bcast_unlock(net);
return 0;
}
tipc: introduce variable window congestion control We introduce a simple variable window congestion control for links. The algorithm is inspired by the Reno algorithm, covering both 'slow start', 'congestion avoidance', and 'fast recovery' modes. - We introduce hard lower and upper window limits per link, still different and configurable per bearer type. - We introduce a 'slow start theshold' variable, initially set to the maximum window size. - We let a link start at the minimum congestion window, i.e. in slow start mode, and then let is grow rapidly (+1 per rceived ACK) until it reaches the slow start threshold and enters congestion avoidance mode. - In congestion avoidance mode we increment the congestion window for each window-size number of acked packets, up to a possible maximum equal to the configured maximum window. - For each non-duplicate NACK received, we drop back to fast recovery mode, by setting the both the slow start threshold to and the congestion window to (current_congestion_window / 2). - If the timeout handler finds that the transmit queue has not moved since the previous timeout, it drops the link back to slow start and forces a probe containing the last sent sequence number to the sent to the peer, so that this can discover the stale situation. This change does in reality have effect only on unicast ethernet transport, as we have seen that there is no room whatsoever for increasing the window max size for the UDP bearer. For now, we also choose to keep the limits for the broadcast link unchanged and equal. This algorithm seems to give a 50-100% throughput improvement for messages larger than MTU. Suggested-by: Xin Long <lucien.xin@gmail.com> 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>
2019-12-10 06:52:46 +07:00
static int tipc_bc_link_set_queue_limits(struct net *net, u32 max_win)
{
struct tipc_link *l = tipc_bc_sndlink(net);
if (!l)
return -ENOPROTOOPT;
tipc: introduce variable window congestion control We introduce a simple variable window congestion control for links. The algorithm is inspired by the Reno algorithm, covering both 'slow start', 'congestion avoidance', and 'fast recovery' modes. - We introduce hard lower and upper window limits per link, still different and configurable per bearer type. - We introduce a 'slow start theshold' variable, initially set to the maximum window size. - We let a link start at the minimum congestion window, i.e. in slow start mode, and then let is grow rapidly (+1 per rceived ACK) until it reaches the slow start threshold and enters congestion avoidance mode. - In congestion avoidance mode we increment the congestion window for each window-size number of acked packets, up to a possible maximum equal to the configured maximum window. - For each non-duplicate NACK received, we drop back to fast recovery mode, by setting the both the slow start threshold to and the congestion window to (current_congestion_window / 2). - If the timeout handler finds that the transmit queue has not moved since the previous timeout, it drops the link back to slow start and forces a probe containing the last sent sequence number to the sent to the peer, so that this can discover the stale situation. This change does in reality have effect only on unicast ethernet transport, as we have seen that there is no room whatsoever for increasing the window max size for the UDP bearer. For now, we also choose to keep the limits for the broadcast link unchanged and equal. This algorithm seems to give a 50-100% throughput improvement for messages larger than MTU. Suggested-by: Xin Long <lucien.xin@gmail.com> 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>
2019-12-10 06:52:46 +07:00
if (max_win < BCLINK_WIN_MIN)
max_win = BCLINK_WIN_MIN;
if (max_win > TIPC_MAX_LINK_WIN)
return -EINVAL;
tipc_bcast_lock(net);
tipc: introduce variable window congestion control We introduce a simple variable window congestion control for links. The algorithm is inspired by the Reno algorithm, covering both 'slow start', 'congestion avoidance', and 'fast recovery' modes. - We introduce hard lower and upper window limits per link, still different and configurable per bearer type. - We introduce a 'slow start theshold' variable, initially set to the maximum window size. - We let a link start at the minimum congestion window, i.e. in slow start mode, and then let is grow rapidly (+1 per rceived ACK) until it reaches the slow start threshold and enters congestion avoidance mode. - In congestion avoidance mode we increment the congestion window for each window-size number of acked packets, up to a possible maximum equal to the configured maximum window. - For each non-duplicate NACK received, we drop back to fast recovery mode, by setting the both the slow start threshold to and the congestion window to (current_congestion_window / 2). - If the timeout handler finds that the transmit queue has not moved since the previous timeout, it drops the link back to slow start and forces a probe containing the last sent sequence number to the sent to the peer, so that this can discover the stale situation. This change does in reality have effect only on unicast ethernet transport, as we have seen that there is no room whatsoever for increasing the window max size for the UDP bearer. For now, we also choose to keep the limits for the broadcast link unchanged and equal. This algorithm seems to give a 50-100% throughput improvement for messages larger than MTU. Suggested-by: Xin Long <lucien.xin@gmail.com> 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>
2019-12-10 06:52:46 +07:00
tipc_link_set_queue_limits(l, BCLINK_WIN_MIN, max_win);
tipc_bcast_unlock(net);
return 0;
}
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
static int tipc_bc_link_set_broadcast_mode(struct net *net, u32 bc_mode)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
switch (bc_mode) {
case BCLINK_MODE_BCAST:
if (!bb->bcast_support)
return -ENOPROTOOPT;
bb->force_bcast = true;
bb->force_rcast = false;
break;
case BCLINK_MODE_RCAST:
if (!bb->rcast_support)
return -ENOPROTOOPT;
bb->force_bcast = false;
bb->force_rcast = true;
break;
case BCLINK_MODE_SEL:
if (!bb->bcast_support || !bb->rcast_support)
return -ENOPROTOOPT;
bb->force_bcast = false;
bb->force_rcast = false;
break;
default:
return -EINVAL;
}
return 0;
}
static int tipc_bc_link_set_broadcast_ratio(struct net *net, u32 bc_ratio)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
if (!bb->bcast_support || !bb->rcast_support)
return -ENOPROTOOPT;
if (bc_ratio > 100 || bc_ratio <= 0)
return -EINVAL;
bb->rc_ratio = bc_ratio;
tipc_bcast_lock(net);
tipc_bcbase_calc_bc_threshold(net);
tipc_bcast_unlock(net);
return 0;
}
int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
{
int err;
u32 win;
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
u32 bc_mode;
u32 bc_ratio;
struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
if (!attrs[TIPC_NLA_LINK_PROP])
return -EINVAL;
err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP], props);
if (err)
return err;
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
if (!props[TIPC_NLA_PROP_WIN] &&
!props[TIPC_NLA_PROP_BROADCAST] &&
!props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
return -EOPNOTSUPP;
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
}
if (props[TIPC_NLA_PROP_BROADCAST]) {
bc_mode = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST]);
err = tipc_bc_link_set_broadcast_mode(net, bc_mode);
}
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
if (!err && props[TIPC_NLA_PROP_BROADCAST_RATIO]) {
bc_ratio = nla_get_u32(props[TIPC_NLA_PROP_BROADCAST_RATIO]);
err = tipc_bc_link_set_broadcast_ratio(net, bc_ratio);
}
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
if (!err && props[TIPC_NLA_PROP_WIN]) {
win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
err = tipc_bc_link_set_queue_limits(net, win);
}
return err;
}
int tipc_bcast_init(struct net *net)
{
struct tipc_net *tn = tipc_net(net);
struct tipc_bc_base *bb = NULL;
struct tipc_link *l = NULL;
bb = kzalloc(sizeof(*bb), GFP_KERNEL);
if (!bb)
goto enomem;
tn->bcbase = bb;
spin_lock_init(&tipc_net(net)->bclock);
if (!tipc_link_bc_create(net, 0, 0,
FB_MTU,
BCLINK_WIN_DEFAULT,
tipc: introduce variable window congestion control We introduce a simple variable window congestion control for links. The algorithm is inspired by the Reno algorithm, covering both 'slow start', 'congestion avoidance', and 'fast recovery' modes. - We introduce hard lower and upper window limits per link, still different and configurable per bearer type. - We introduce a 'slow start theshold' variable, initially set to the maximum window size. - We let a link start at the minimum congestion window, i.e. in slow start mode, and then let is grow rapidly (+1 per rceived ACK) until it reaches the slow start threshold and enters congestion avoidance mode. - In congestion avoidance mode we increment the congestion window for each window-size number of acked packets, up to a possible maximum equal to the configured maximum window. - For each non-duplicate NACK received, we drop back to fast recovery mode, by setting the both the slow start threshold to and the congestion window to (current_congestion_window / 2). - If the timeout handler finds that the transmit queue has not moved since the previous timeout, it drops the link back to slow start and forces a probe containing the last sent sequence number to the sent to the peer, so that this can discover the stale situation. This change does in reality have effect only on unicast ethernet transport, as we have seen that there is no room whatsoever for increasing the window max size for the UDP bearer. For now, we also choose to keep the limits for the broadcast link unchanged and equal. This algorithm seems to give a 50-100% throughput improvement for messages larger than MTU. Suggested-by: Xin Long <lucien.xin@gmail.com> 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>
2019-12-10 06:52:46 +07:00
BCLINK_WIN_DEFAULT,
0,
&bb->inputq,
NULL,
NULL,
&l))
goto enomem;
bb->link = l;
tn->bcl = l;
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
bb->rc_ratio = 10;
bb->rcast_support = true;
return 0;
enomem:
kfree(bb);
kfree(l);
return -ENOMEM;
}
void tipc_bcast_stop(struct net *net)
{
struct tipc_net *tn = net_generic(net, tipc_net_id);
synchronize_net();
kfree(tn->bcbase);
kfree(tn->bcl);
}
void tipc_nlist_init(struct tipc_nlist *nl, u32 self)
{
memset(nl, 0, sizeof(*nl));
INIT_LIST_HEAD(&nl->list);
nl->self = self;
}
void tipc_nlist_add(struct tipc_nlist *nl, u32 node)
{
if (node == nl->self)
nl->local = true;
else if (tipc_dest_push(&nl->list, node, 0))
nl->remote++;
}
void tipc_nlist_del(struct tipc_nlist *nl, u32 node)
{
if (node == nl->self)
nl->local = false;
else if (tipc_dest_del(&nl->list, node, 0))
nl->remote--;
}
void tipc_nlist_purge(struct tipc_nlist *nl)
{
tipc_dest_list_purge(&nl->list);
nl->remote = 0;
nl->local = false;
}
tipc: support broadcast/replicast configurable for bc-link Currently, a multicast stream uses either broadcast or replicast as transmission method, based on the ratio between number of actual destinations nodes and cluster size. However, when an L2 interface (e.g., VXLAN) provides pseudo broadcast support, this becomes very inefficient, as it blindly replicates multicast packets to all cluster/subnet nodes, irrespective of whether they host actual target sockets or not. The TIPC multicast algorithm is able to distinguish real destination nodes from other nodes, and hence provides a smarter and more efficient method for transferring multicast messages than pseudo broadcast can do. Because of this, we now make it possible for users to force the broadcast link to permanently switch to using replicast, irrespective of which capabilities the bearer provides, or pretend to provide. Conversely, we also make it possible to force the broadcast link to always use true broadcast. While maybe less useful in deployed systems, this may at least be useful for testing the broadcast algorithm in small clusters. We retain the current AUTOSELECT ability, i.e., to let the broadcast link automatically select which algorithm to use, and to switch back and forth between broadcast and replicast as the ratio between destination node number and cluster size changes. This remains the default method. Furthermore, we make it possible to configure the threshold ratio for such switches. The default ratio is now set to 10%, down from 25% in the earlier implementation. Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-19 18:49:48 +07:00
u32 tipc_bcast_get_broadcast_mode(struct net *net)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
if (bb->force_bcast)
return BCLINK_MODE_BCAST;
if (bb->force_rcast)
return BCLINK_MODE_RCAST;
if (bb->bcast_support && bb->rcast_support)
return BCLINK_MODE_SEL;
return 0;
}
u32 tipc_bcast_get_broadcast_ratio(struct net *net)
{
struct tipc_bc_base *bb = tipc_bc_base(net);
return bb->rc_ratio;
}
tipc: fix a null pointer deref In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast") we introduced new method to eliminate the risk of message reordering that happen in between different nodes. Unfortunately, we forgot checking at receiving side to ignore intra node. We fix this by checking and returning if arrived message from intra node. syzbot report: ================================================================== kasan: CONFIG_KASAN_INLINE enabled kasan: GPF could be caused by NULL-ptr deref or user memory access general protection fault: 0000 [#1] PREEMPT SMP KASAN CPU: 0 PID: 7820 Comm: syz-executor418 Not tainted 5.0.0+ #61 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:tipc_mcast_filter_msg+0x21b/0x13d0 net/tipc/bcast.c:782 Code: 45 c0 0f 84 39 06 00 00 48 89 5d 98 e8 ce ab a5 fa 49 8d bc 24 c8 00 00 00 48 b9 00 00 00 00 00 fc ff df 48 89 f8 48 c1 e8 03 <80> 3c 08 00 0f 85 9a 0e 00 00 49 8b 9c 24 c8 00 00 00 48 be 00 00 RSP: 0018:ffff8880959defc8 EFLAGS: 00010202 RAX: 0000000000000019 RBX: ffff888081258a48 RCX: dffffc0000000000 RDX: 0000000000000000 RSI: ffffffff86cab862 RDI: 00000000000000c8 RBP: ffff8880959df030 R08: ffff8880813d0200 R09: ffffed1015d05bc8 R10: ffffed1015d05bc7 R11: ffff8880ae82de3b R12: 0000000000000000 R13: 000000000000002c R14: 0000000000000000 R15: ffff888081258a48 FS: 000000000106a880(0000) GS:ffff8880ae800000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000020001cc0 CR3: 0000000094a20000 CR4: 00000000001406f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: tipc_sk_filter_rcv+0x182d/0x34f0 net/tipc/socket.c:2168 tipc_sk_enqueue net/tipc/socket.c:2254 [inline] tipc_sk_rcv+0xc45/0x25a0 net/tipc/socket.c:2305 tipc_sk_mcast_rcv+0x724/0x1020 net/tipc/socket.c:1209 tipc_mcast_xmit+0x7fe/0x1200 net/tipc/bcast.c:410 tipc_sendmcast+0xb36/0xfc0 net/tipc/socket.c:820 __tipc_sendmsg+0x10df/0x18d0 net/tipc/socket.c:1358 tipc_sendmsg+0x53/0x80 net/tipc/socket.c:1291 sock_sendmsg_nosec net/socket.c:651 [inline] sock_sendmsg+0xdd/0x130 net/socket.c:661 ___sys_sendmsg+0x806/0x930 net/socket.c:2260 __sys_sendmsg+0x105/0x1d0 net/socket.c:2298 __do_sys_sendmsg net/socket.c:2307 [inline] __se_sys_sendmsg net/socket.c:2305 [inline] __x64_sys_sendmsg+0x78/0xb0 net/socket.c:2305 do_syscall_64+0x103/0x610 arch/x86/entry/common.c:290 entry_SYSCALL_64_after_hwframe+0x49/0xbe RIP: 0033:0x4401c9 Code: 18 89 d0 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 0f 83 fb 13 fc ff c3 66 2e 0f 1f 84 00 00 00 00 RSP: 002b:00007ffd887fa9d8 EFLAGS: 00000246 ORIG_RAX: 000000000000002e RAX: ffffffffffffffda RBX: 00000000004002c8 RCX: 00000000004401c9 RDX: 0000000000000000 RSI: 0000000020002140 RDI: 0000000000000003 RBP: 00000000006ca018 R08: 0000000000000000 R09: 00000000004002c8 R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000401a50 R13: 0000000000401ae0 R14: 0000000000000000 R15: 0000000000000000 Modules linked in: ---[ end trace ba79875754e1708f ]--- Reported-by: syzbot+be4bdf2cc3e85e952c50@syzkaller.appspotmail.com Fixes: c55c8eda ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-21 17:25:18 +07:00
void tipc_mcast_filter_msg(struct net *net, struct sk_buff_head *defq,
struct sk_buff_head *inputq)
{
struct sk_buff *skb, *_skb, *tmp;
struct tipc_msg *hdr, *_hdr;
bool match = false;
u32 node, port;
skb = skb_peek(inputq);
if (!skb)
return;
hdr = buf_msg(skb);
if (likely(!msg_is_syn(hdr) && skb_queue_empty(defq)))
return;
node = msg_orignode(hdr);
tipc: fix a null pointer deref In commit c55c8edafa91 ("tipc: smooth change between replicast and broadcast") we introduced new method to eliminate the risk of message reordering that happen in between different nodes. Unfortunately, we forgot checking at receiving side to ignore intra node. We fix this by checking and returning if arrived message from intra node. syzbot report: ================================================================== kasan: CONFIG_KASAN_INLINE enabled kasan: GPF could be caused by NULL-ptr deref or user memory access general protection fault: 0000 [#1] PREEMPT SMP KASAN CPU: 0 PID: 7820 Comm: syz-executor418 Not tainted 5.0.0+ #61 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 RIP: 0010:tipc_mcast_filter_msg+0x21b/0x13d0 net/tipc/bcast.c:782 Code: 45 c0 0f 84 39 06 00 00 48 89 5d 98 e8 ce ab a5 fa 49 8d bc 24 c8 00 00 00 48 b9 00 00 00 00 00 fc ff df 48 89 f8 48 c1 e8 03 <80> 3c 08 00 0f 85 9a 0e 00 00 49 8b 9c 24 c8 00 00 00 48 be 00 00 RSP: 0018:ffff8880959defc8 EFLAGS: 00010202 RAX: 0000000000000019 RBX: ffff888081258a48 RCX: dffffc0000000000 RDX: 0000000000000000 RSI: ffffffff86cab862 RDI: 00000000000000c8 RBP: ffff8880959df030 R08: ffff8880813d0200 R09: ffffed1015d05bc8 R10: ffffed1015d05bc7 R11: ffff8880ae82de3b R12: 0000000000000000 R13: 000000000000002c R14: 0000000000000000 R15: ffff888081258a48 FS: 000000000106a880(0000) GS:ffff8880ae800000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000020001cc0 CR3: 0000000094a20000 CR4: 00000000001406f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: tipc_sk_filter_rcv+0x182d/0x34f0 net/tipc/socket.c:2168 tipc_sk_enqueue net/tipc/socket.c:2254 [inline] tipc_sk_rcv+0xc45/0x25a0 net/tipc/socket.c:2305 tipc_sk_mcast_rcv+0x724/0x1020 net/tipc/socket.c:1209 tipc_mcast_xmit+0x7fe/0x1200 net/tipc/bcast.c:410 tipc_sendmcast+0xb36/0xfc0 net/tipc/socket.c:820 __tipc_sendmsg+0x10df/0x18d0 net/tipc/socket.c:1358 tipc_sendmsg+0x53/0x80 net/tipc/socket.c:1291 sock_sendmsg_nosec net/socket.c:651 [inline] sock_sendmsg+0xdd/0x130 net/socket.c:661 ___sys_sendmsg+0x806/0x930 net/socket.c:2260 __sys_sendmsg+0x105/0x1d0 net/socket.c:2298 __do_sys_sendmsg net/socket.c:2307 [inline] __se_sys_sendmsg net/socket.c:2305 [inline] __x64_sys_sendmsg+0x78/0xb0 net/socket.c:2305 do_syscall_64+0x103/0x610 arch/x86/entry/common.c:290 entry_SYSCALL_64_after_hwframe+0x49/0xbe RIP: 0033:0x4401c9 Code: 18 89 d0 c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 0f 83 fb 13 fc ff c3 66 2e 0f 1f 84 00 00 00 00 RSP: 002b:00007ffd887fa9d8 EFLAGS: 00000246 ORIG_RAX: 000000000000002e RAX: ffffffffffffffda RBX: 00000000004002c8 RCX: 00000000004401c9 RDX: 0000000000000000 RSI: 0000000020002140 RDI: 0000000000000003 RBP: 00000000006ca018 R08: 0000000000000000 R09: 00000000004002c8 R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000401a50 R13: 0000000000401ae0 R14: 0000000000000000 R15: 0000000000000000 Modules linked in: ---[ end trace ba79875754e1708f ]--- Reported-by: syzbot+be4bdf2cc3e85e952c50@syzkaller.appspotmail.com Fixes: c55c8eda ("tipc: smooth change between replicast and broadcast") Acked-by: Jon Maloy <jon.maloy@ericsson.com> Signed-off-by: Hoang Le <hoang.h.le@dektech.com.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-21 17:25:18 +07:00
if (node == tipc_own_addr(net))
return;
port = msg_origport(hdr);
/* Has the twin SYN message already arrived ? */
skb_queue_walk(defq, _skb) {
_hdr = buf_msg(_skb);
if (msg_orignode(_hdr) != node)
continue;
if (msg_origport(_hdr) != port)
continue;
match = true;
break;
}
if (!match) {
if (!msg_is_syn(hdr))
return;
__skb_dequeue(inputq);
__skb_queue_tail(defq, skb);
return;
}
/* Deliver non-SYN message from other link, otherwise queue it */
if (!msg_is_syn(hdr)) {
if (msg_is_rcast(hdr) != msg_is_rcast(_hdr))
return;
__skb_dequeue(inputq);
__skb_queue_tail(defq, skb);
return;
}
/* Queue non-SYN/SYN message from same link */
if (msg_is_rcast(hdr) == msg_is_rcast(_hdr)) {
__skb_dequeue(inputq);
__skb_queue_tail(defq, skb);
return;
}
/* Matching SYN messages => return the one with data, if any */
__skb_unlink(_skb, defq);
if (msg_data_sz(hdr)) {
kfree_skb(_skb);
} else {
__skb_dequeue(inputq);
kfree_skb(skb);
__skb_queue_tail(inputq, _skb);
}
/* Deliver subsequent non-SYN messages from same peer */
skb_queue_walk_safe(defq, _skb, tmp) {
_hdr = buf_msg(_skb);
if (msg_orignode(_hdr) != node)
continue;
if (msg_origport(_hdr) != port)
continue;
if (msg_is_syn(_hdr))
break;
__skb_unlink(_skb, defq);
__skb_queue_tail(inputq, _skb);
}
}