linux/net/rds/cong.c

/*
 * Copyright (c) 2007, 2017 Oracle and/or its affiliates. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */
#include <linux/slab.h>
#include <linux/types.h>
#include <linux/rbtree.h>
#include <linux/bitops.h>
#include <linux/export.h>

#include "rds.h"

/*
 * This file implements the receive side of the unconventional congestion
 * management in RDS.
 *
 * Messages waiting in the receive queue on the receiving socket are accounted
 * against the sockets SO_RCVBUF option value.  Only the payload bytes in the
 * message are accounted for.  If the number of bytes queued equals or exceeds
 * rcvbuf then the socket is congested.  All sends attempted to this socket's
 * address should return block or return -EWOULDBLOCK.
 *
 * Applications are expected to be reasonably tuned such that this situation
 * very rarely occurs.  An application encountering this "back-pressure" is
 * considered a bug.
 *
 * This is implemented by having each node maintain bitmaps which indicate
 * which ports on bound addresses are congested.  As the bitmap changes it is
 * sent through all the connections which terminate in the local address of the
 * bitmap which changed.
 *
 * The bitmaps are allocated as connections are brought up.  This avoids
 * allocation in the interrupt handling path which queues messages on sockets.
 * The dense bitmaps let transports send the entire bitmap on any bitmap change
 * reasonably efficiently.  This is much easier to implement than some
 * finer-grained communication of per-port congestion.  The sender does a very
 * inexpensive bit test to test if the port it's about to send to is congested
 * or not.
 */

/*
 * Interaction with poll is a tad tricky. We want all processes stuck in
 * poll to wake up and check whether a congested destination became uncongested.
 * The really sad thing is we have no idea which destinations the application
 * wants to send to - we don't even know which rds_connections are involved.
 * So until we implement a more flexible rds poll interface, we have to make
 * do with this:
 * We maintain a global counter that is incremented each time a congestion map
 * update is received. Each rds socket tracks this value, and if rds_poll
 * finds that the saved generation number is smaller than the global generation
 * number, it wakes up the process.
 */
static atomic_t		rds_cong_generation = …;

/*
 * Congestion monitoring
 */
static LIST_HEAD(rds_cong_monitor);
static DEFINE_RWLOCK(rds_cong_monitor_lock);

/*
 * Yes, a global lock.  It's used so infrequently that it's worth keeping it
 * global to simplify the locking.  It's only used in the following
 * circumstances:
 *
 *  - on connection buildup to associate a conn with its maps
 *  - on map changes to inform conns of a new map to send
 *
 *  It's sadly ordered under the socket callback lock and the connection lock.
 *  Receive paths can mark ports congested from interrupt context so the
 *  lock masks interrupts.
 */
static DEFINE_SPINLOCK(rds_cong_lock);
static struct rb_root rds_cong_tree = …;

static struct rds_cong_map *rds_cong_tree_walk(const struct in6_addr *addr,
					       struct rds_cong_map *insert)
{ … }

/*
 * There is only ever one bitmap for any address.  Connections try and allocate
 * these bitmaps in the process getting pointers to them.  The bitmaps are only
 * ever freed as the module is removed after all connections have been freed.
 */
static struct rds_cong_map *rds_cong_from_addr(const struct in6_addr *addr)
{ … }

/*
 * Put the conn on its local map's list.  This is called when the conn is
 * really added to the hash.  It's nested under the rds_conn_lock, sadly.
 */
void rds_cong_add_conn(struct rds_connection *conn)
{ … }

void rds_cong_remove_conn(struct rds_connection *conn)
{ … }

int rds_cong_get_maps(struct rds_connection *conn)
{ … }

void rds_cong_queue_updates(struct rds_cong_map *map)
{ … }

void rds_cong_map_updated(struct rds_cong_map *map, uint64_t portmask)
{ … }
EXPORT_SYMBOL_GPL(…);

int rds_cong_updated_since(unsigned long *recent)
{ … }

/*
 * We're called under the locking that protects the sockets receive buffer
 * consumption.  This makes it a lot easier for the caller to only call us
 * when it knows that an existing set bit needs to be cleared, and vice versa.
 * We can't block and we need to deal with concurrent sockets working against
 * the same per-address map.
 */
void rds_cong_set_bit(struct rds_cong_map *map, __be16 port)
{ … }

void rds_cong_clear_bit(struct rds_cong_map *map, __be16 port)
{ … }

static int rds_cong_test_bit(struct rds_cong_map *map, __be16 port)
{ … }

void rds_cong_add_socket(struct rds_sock *rs)
{ … }

void rds_cong_remove_socket(struct rds_sock *rs)
{ … }

int rds_cong_wait(struct rds_cong_map *map, __be16 port, int nonblock,
		  struct rds_sock *rs)
{ … }

void rds_cong_exit(void)
{ … }

/*
 * Allocate a RDS message containing a congestion update.
 */
struct rds_message *rds_cong_update_alloc(struct rds_connection *conn)
{ … }