linux/net/tipc/bcast.c
Jon Maloy 4c94cc2d3d tipc: fall back to smaller MTU if allocation of local send skb fails
When sending node local messages the code is using an 'mtu' of 66060
bytes to avoid unnecessary fragmentation. During situations of low
memory tipc_msg_build() may sometimes fail to allocate such large
buffers, resulting in unnecessary send failures. This can easily be
remedied by falling back to a smaller MTU, and then reassemble the
buffer chain as if the message were arriving from a remote node.

At the same time, we change the initial MTU setting of the broadcast
link to a lower value, so that large messages always are fragmented
into smaller buffers even when we run in single node mode. Apart from
obtaining the same advantage as for the 'fallback' solution above, this
turns out to give a significant performance improvement. This can
probably be explained with the __pskb_copy() operation performed on the
buffer for each recipient during reception. We found the optimal value
for this, considering the most relevant skb pool, to be 3744 bytes.

Acked-by: Ying Xue <ying.xue@ericsson.com>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2017-12-01 15:21:25 -05:00

579 lines
16 KiB
C

/*
* 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
* @dest: 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
* @rcast_support: indicates if all peer nodes support replicast
* @rc_ratio: dest count as percentage of cluster size where send method changes
* @bc_threshold: calculated drom 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;
bool rcast_support;
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)
{
return tipc_link_mtu(tipc_bc_sndlink(net)) - INT_H_SIZE;
}
void tipc_bcast_disable_rcast(struct net *net)
{
tipc_bc_base(net)->rcast_support = false;
}
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 */
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 || time_before(jiffies, exp))
return;
/* 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
* 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;
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)));
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_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;
int rc = 0;
skb_queue_head_init(&inputq);
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);
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_sk_mcast_rcv(net, &localq, &inputq);
exit:
/* This queue should normally be empty by now */
__skb_queue_purge(pkts);
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
*/
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;
u16 acked = msg_bcast_ack(hdr);
struct sk_buff_head xmitq;
/* 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);
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;
}
static int tipc_bc_link_set_queue_limits(struct net *net, u32 limit)
{
struct tipc_link *l = tipc_bc_sndlink(net);
if (!l)
return -ENOPROTOOPT;
if (limit < BCLINK_WIN_MIN)
limit = BCLINK_WIN_MIN;
if (limit > TIPC_MAX_LINK_WIN)
return -EINVAL;
tipc_bcast_lock(net);
tipc_link_set_queue_limits(l, limit);
tipc_bcast_unlock(net);
return 0;
}
int tipc_nl_bc_link_set(struct net *net, struct nlattr *attrs[])
{
int err;
u32 win;
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;
if (!props[TIPC_NLA_PROP_WIN])
return -EOPNOTSUPP;
win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
return tipc_bc_link_set_queue_limits(net, win);
}
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_ATOMIC);
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,
0,
&bb->inputq,
NULL,
NULL,
&l))
goto enomem;
bb->link = l;
tn->bcl = l;
bb->rc_ratio = 25;
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 = 0;
}