clients/rdsv3/cong.c

/*
 * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved.
 */

/*
 * This file contains code imported from the OFED rds source file cong.c
 * Oracle elects to have and use the contents of cong.c under and governed
 * by the OpenIB.org BSD license (see below for full license text). However,
 * the following notice accompanied the original version of this file:
 */


/*
 * Copyright (c) 2007 Oracle.  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 <sys/rds.h>

#include <sys/ib/clients/rdsv3/rdsv3.h>
#include <sys/ib/clients/rdsv3/rdsv3_impl.h>
#include <sys/ib/clients/rdsv3/rdsv3_debug.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 rdsv3_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 rdsv3_poll
 * finds that the saved generation number is smaller than the global generation
 * number, it wakes up the process.
 */
static atomic_t		rdsv3_cong_generation = ATOMIC_INIT(0);

/*
 * Congestion monitoring
 */
static struct list rdsv3_cong_monitor;
static krwlock_t rdsv3_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 kmutex_t rdsv3_cong_lock;
static struct avl_tree rdsv3_cong_tree;

static struct rdsv3_cong_map *
rdsv3_cong_tree_walk(uint32_be_t addr, struct rdsv3_cong_map *insert)
{
	struct rdsv3_cong_map *map;
	avl_index_t where;

	if (insert) {
		map = avl_find(&rdsv3_cong_tree, insert, &where);
		if (map == NULL) {
			avl_insert(&rdsv3_cong_tree, insert, where);
			return (NULL);
		}
	} else {
		struct rdsv3_cong_map map1;
		map1.m_addr = addr;
		map = avl_find(&rdsv3_cong_tree, &map1, &where);
	}

	return (map);
}

/*
 * 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 rdsv3_cong_map *
rdsv3_cong_from_addr(uint32_be_t addr)
{
	struct rdsv3_cong_map *map;
	struct rdsv3_cong_map *ret = NULL;
	unsigned long zp;
	unsigned long i;

	RDSV3_DPRINTF4("rdsv3_cong_from_addr", "Enter(addr: %x)", ntohl(addr));

	map = kmem_zalloc(sizeof (struct rdsv3_cong_map), KM_NOSLEEP);
	if (!map)
		return (NULL);

	map->m_addr = addr;
	rdsv3_init_waitqueue(&map->m_waitq);
	list_create(&map->m_conn_list, sizeof (struct rdsv3_connection),
	    offsetof(struct rdsv3_connection, c_map_item));

	for (i = 0; i < RDSV3_CONG_MAP_PAGES; i++) {
		zp = (unsigned long)kmem_zalloc(PAGE_SIZE, KM_NOSLEEP);
		if (zp == 0)
			goto out;
		map->m_page_addrs[i] = zp;
	}

	mutex_enter(&rdsv3_cong_lock);
	ret = rdsv3_cong_tree_walk(addr, map);
	mutex_exit(&rdsv3_cong_lock);

	if (!ret) {
		ret = map;
		map = NULL;
	}

out:
	if (map) {
		for (i = 0; i < RDSV3_CONG_MAP_PAGES && map->m_page_addrs[i];
		    i++)
			kmem_free((void *)map->m_page_addrs[i], PAGE_SIZE);
		kmem_free(map, sizeof (*map));
	}

	RDSV3_DPRINTF5("rdsv3_cong_from_addr", "map %p for addr %x",
	    ret, ntohl(addr));

	return (ret);
}

/*
 * 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 rdsv3_conn_lock, sadly.
 */
void
rdsv3_cong_add_conn(struct rdsv3_connection *conn)
{
	RDSV3_DPRINTF4("rdsv3_cong_add_conn", "Enter(conn: %p)", conn);

	RDSV3_DPRINTF5("rdsv3_cong_add_conn", "conn %p now on map %p",
	    conn, conn->c_lcong);
	mutex_enter(&rdsv3_cong_lock);
	list_insert_tail(&conn->c_lcong->m_conn_list, conn);
	mutex_exit(&rdsv3_cong_lock);

	RDSV3_DPRINTF4("rdsv3_cong_add_conn", "Return(conn: %p)", conn);
}

void
rdsv3_cong_remove_conn(struct rdsv3_connection *conn)
{
	RDSV3_DPRINTF4("rdsv3_cong_remove_conn", "Enter(conn: %p)", conn);

	RDSV3_DPRINTF5("rdsv3_cong_remove_conn", "removing conn %p from map %p",
	    conn, conn->c_lcong);
	mutex_enter(&rdsv3_cong_lock);
	list_remove_node(&conn->c_map_item);
	mutex_exit(&rdsv3_cong_lock);

	RDSV3_DPRINTF4("rdsv3_cong_remove_conn", "Return(conn: %p)", conn);
}

int
rdsv3_cong_get_maps(struct rdsv3_connection *conn)
{
	conn->c_lcong = rdsv3_cong_from_addr(conn->c_laddr);
	conn->c_fcong = rdsv3_cong_from_addr(conn->c_faddr);

	if (!(conn->c_lcong && conn->c_fcong))
		return (-ENOMEM);

	return (0);
}

void
rdsv3_cong_queue_updates(struct rdsv3_cong_map *map)
{
	struct rdsv3_connection *conn;

	RDSV3_DPRINTF4("rdsv3_cong_queue_updates", "Enter(map: %p)", map);

	mutex_enter(&rdsv3_cong_lock);

	RDSV3_FOR_EACH_LIST_NODE(conn, &map->m_conn_list, c_map_item) {
		if (!test_and_set_bit(0, &conn->c_map_queued)) {
			rdsv3_stats_inc(s_cong_update_queued);
			(void) rdsv3_send_xmit(conn);
		}
	}

	mutex_exit(&rdsv3_cong_lock);

	RDSV3_DPRINTF4("rdsv3_cong_queue_updates", "Return(map: %p)", map);
}

void
rdsv3_cong_map_updated(struct rdsv3_cong_map *map, uint64_t portmask)
{
	RDSV3_DPRINTF4("rdsv3_cong_map_updated",
	    "waking map %p for %u.%u.%u.%u",
	    map, NIPQUAD(map->m_addr));

	rdsv3_stats_inc(s_cong_update_received);
	atomic_inc_32(&rdsv3_cong_generation);
#if 0
XXX
	if (waitqueue_active(&map->m_waitq))
#endif
		rdsv3_wake_up(&map->m_waitq);

	if (portmask && !list_is_empty(&rdsv3_cong_monitor)) {
		struct rdsv3_sock *rs;

		rw_enter(&rdsv3_cong_monitor_lock, RW_READER);
		RDSV3_FOR_EACH_LIST_NODE(rs, &rdsv3_cong_monitor,
		    rs_cong_list) {
			mutex_enter(&rs->rs_lock);
			rs->rs_cong_notify |= (rs->rs_cong_mask & portmask);
			rs->rs_cong_mask &= ~portmask;
			mutex_exit(&rs->rs_lock);
			if (rs->rs_cong_notify)
				rdsv3_wake_sk_sleep(rs);
		}
		rw_exit(&rdsv3_cong_monitor_lock);
	}

	RDSV3_DPRINTF4("rdsv3_cong_map_updated", "Return(map: %p)", map);
}

int
rdsv3_cong_updated_since(unsigned long *recent)
{
	unsigned long gen = atomic_get(&rdsv3_cong_generation);

	if (*recent == gen)
		return (0);
	*recent = gen;
	return (1);
}

/*
 * 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
rdsv3_cong_set_bit(struct rdsv3_cong_map *map, uint16_be_t port)
{
	unsigned long i;
	unsigned long off;

	RDSV3_DPRINTF4("rdsv3_cong_set_bit",
	    "setting congestion for %u.%u.%u.%u:%u in map %p",
	    NIPQUAD(map->m_addr), ntohs(port), map);

	i = ntohs(port) / RDSV3_CONG_MAP_PAGE_BITS;
	off = ntohs(port) % RDSV3_CONG_MAP_PAGE_BITS;
	set_le_bit(off, (void *)map->m_page_addrs[i]);
}

void
rdsv3_cong_clear_bit(struct rdsv3_cong_map *map, uint16_be_t port)
{
	unsigned long i;
	unsigned long off;

	RDSV3_DPRINTF4("rdsv3_cong_clear_bit",
	    "clearing congestion for %u.%u.%u.%u:%u in map %p\n",
	    NIPQUAD(map->m_addr), ntohs(port), map);

	i = ntohs(port) / RDSV3_CONG_MAP_PAGE_BITS;
	off = ntohs(port) % RDSV3_CONG_MAP_PAGE_BITS;
	clear_le_bit(off, (void *)map->m_page_addrs[i]);
}

static int
rdsv3_cong_test_bit(struct rdsv3_cong_map *map, uint16_be_t port)
{
	unsigned long i;
	unsigned long off;

	i = ntohs(port) / RDSV3_CONG_MAP_PAGE_BITS;
	off = ntohs(port) % RDSV3_CONG_MAP_PAGE_BITS;

	RDSV3_DPRINTF5("rdsv3_cong_test_bit", "port: 0x%x i = %lx off = %lx",
	    ntohs(port), i, off);

	return (test_le_bit(off, (void *)map->m_page_addrs[i]));
}

void
rdsv3_cong_add_socket(struct rdsv3_sock *rs)
{
	RDSV3_DPRINTF4("rdsv3_cong_add_socket", "Enter(rs: %p)", rs);

	rw_enter(&rdsv3_cong_monitor_lock, RW_WRITER);
	if (!list_link_active(&rs->rs_cong_list))
		list_insert_head(&rdsv3_cong_monitor, rs);
	rw_exit(&rdsv3_cong_monitor_lock);
}

void
rdsv3_cong_remove_socket(struct rdsv3_sock *rs)
{
	struct rdsv3_cong_map *map;

	RDSV3_DPRINTF4("rdsv3_cong_remove_socket", "Enter(rs: %p)", rs);

	rw_enter(&rdsv3_cong_monitor_lock, RW_WRITER);
	list_remove_node(&rs->rs_cong_list);
	rw_exit(&rdsv3_cong_monitor_lock);

	/* update congestion map for now-closed port */
	mutex_enter(&rdsv3_cong_lock);
	map = rdsv3_cong_tree_walk(rs->rs_bound_addr, NULL);
	mutex_exit(&rdsv3_cong_lock);

	if (map && rdsv3_cong_test_bit(map, rs->rs_bound_port)) {
		rdsv3_cong_clear_bit(map, rs->rs_bound_port);
		rdsv3_cong_queue_updates(map);
	}
}

int
rdsv3_cong_wait(struct rdsv3_cong_map *map, uint16_be_t port, int nonblock,
    struct rdsv3_sock *rs)
{
	int ret = 0;

	RDSV3_DPRINTF4("rdsv3_cong_wait", "Enter(rs: %p, mode: %d)",
	    rs, nonblock);

	if (!rdsv3_cong_test_bit(map, port))
		return (0);
	if (nonblock) {
		if (rs && rs->rs_cong_monitor) {
			/*
			 * It would have been nice to have an atomic set_bit on
			 * a uint64_t.
			 */
			mutex_enter(&rs->rs_lock);
			rs->rs_cong_mask |=
			    RDS_CONG_MONITOR_MASK(ntohs(port));
			mutex_exit(&rs->rs_lock);

			/*
			 * Test again - a congestion update may have arrived in
			 * the meantime.
			 */
			if (!rdsv3_cong_test_bit(map, port))
				return (0);
		}
		rdsv3_stats_inc(s_cong_send_error);
		return (-ENOBUFS);
	}

	rdsv3_stats_inc(s_cong_send_blocked);
	RDSV3_DPRINTF3("rdsv3_cong_wait", "waiting on map %p for port %u",
	    map, ntohs(port));

#if 0
	ret = rdsv3_wait_sig(&map->m_waitq, !rdsv3_cong_test_bit(map, port));
	if (ret == 0)
		return (-ERESTART);
	return (0);
#else
	mutex_enter(&map->m_waitq.waitq_mutex);
	map->m_waitq.waitq_waiters++;
	while (rdsv3_cong_test_bit(map, port)) {
		ret = cv_wait_sig(&map->m_waitq.waitq_cv,
		    &map->m_waitq.waitq_mutex);
		if (ret == 0) {
			ret = -EINTR;
			break;
		}
	}
	map->m_waitq.waitq_waiters--;
	mutex_exit(&map->m_waitq.waitq_mutex);
	return (ret);
#endif
}

void
rdsv3_cong_exit(void)
{
	struct rdsv3_cong_map *map;
	unsigned long i;

	RDSV3_DPRINTF4("rdsv3_cong_exit", "Enter");

	while ((map = avl_first(&rdsv3_cong_tree))) {
		RDSV3_DPRINTF5("rdsv3_cong_exit", "freeing map %p\n", map);
		avl_remove(&rdsv3_cong_tree, map);
		for (i = 0; i < RDSV3_CONG_MAP_PAGES && map->m_page_addrs[i];
		    i++)
			kmem_free((void *)map->m_page_addrs[i], PAGE_SIZE);
		kmem_free(map, sizeof (*map));
	}

	RDSV3_DPRINTF4("rdsv3_cong_exit", "Return");
}

/*
 * Allocate a RDS message containing a congestion update.
 */
struct rdsv3_message *
rdsv3_cong_update_alloc(struct rdsv3_connection *conn)
{
	struct rdsv3_cong_map *map = conn->c_lcong;
	struct rdsv3_message *rm;

	rm = rdsv3_message_map_pages(map->m_page_addrs, RDSV3_CONG_MAP_BYTES);
	if (!IS_ERR(rm))
		rm->m_inc.i_hdr.h_flags = RDSV3_FLAG_CONG_BITMAP;

	return (rm);
}

static int
rdsv3_cong_compare(const void *map1, const void *map2)
{
#define	addr1	((struct rdsv3_cong_map *)map1)->m_addr
#define	addr2	((struct rdsv3_cong_map *)map2)->m_addr

	if (addr1 < addr2)
		return (-1);
	if (addr1 > addr2)
		return (1);
	return (0);
}

void
rdsv3_cong_init(void)
{
	list_create(&rdsv3_cong_monitor, sizeof (struct rdsv3_sock),
	    offsetof(struct rdsv3_sock, rs_cong_list));
	rw_init(&rdsv3_cong_monitor_lock, NULL, RW_DRIVER, NULL);
	mutex_init(&rdsv3_cong_lock, NULL, MUTEX_DRIVER, NULL);
	avl_create(&rdsv3_cong_tree, rdsv3_cong_compare,
	    sizeof (struct rdsv3_cong_map), offsetof(struct rdsv3_cong_map,
	    m_rb_node));
}