mac80211: mesh path table implementation

The mesh path table associates destinations with the next hop to reach them. The
table is a hash of linked lists protected by rcu mechanisms. Every mesh path
contains a lock to protect the mesh path state.

Each outgoing mesh frame requires a look up into this table. Therefore, the
table it has been designed so it is not necessary to hold any lock to find the
appropriate next hop.

If the path is determined to be active within a rcu context we can safely
dereference mpath->next_hop->addr, since it holds a reference to the sta
next_hop. After a mesh path has been set active for the first time it next_hop
must always point to a valid sta.  If this is not possible the mpath must be
deleted or replaced in a RCU safe fashion.

Signed-off-by: Luis Carlos Cobo <luisca@cozybit.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

net/mac80211/mesh_pathtbl.c

@@ -0,0 +1,522 @@
+/*
+ * Copyright (c) 2008 open80211s Ltd.
+ * Author:     Luis Carlos Cobo <luisca@cozybit.com>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/etherdevice.h>
+#include <linux/list.h>
+#include <linux/netdevice.h>
+#include <linux/random.h>
+#include <linux/spinlock.h>
+#include <linux/string.h>
+#include <net/mac80211.h>
+#include "ieee80211_i.h"
+#include "mesh.h"
+
+/* There will be initially 2^INIT_PATHS_SIZE_ORDER buckets */
+#define INIT_PATHS_SIZE_ORDER	2
+
+/* Keep the mean chain length below this constant */
+#define MEAN_CHAIN_LEN		2
+
+#define MPATH_EXPIRED(mpath) ((mpath->flags & MESH_PATH_ACTIVE) && \
+				time_after(jiffies, mpath->exp_time) && \
+				!(mpath->flags & MESH_PATH_FIXED))
+
+struct mpath_node {
+	struct hlist_node list;
+	struct rcu_head rcu;
+	/* This indirection allows two different tables to point to the same
+	 * mesh_path structure, useful when resizing
+	 */
+	struct mesh_path *mpath;
+};
+
+static struct mesh_table *mesh_paths;
+
+/* This lock will have the grow table function as writer and add / delete nodes
+ * as readers. When reading the table (i.e. doing lookups) we are well protected
+ * by RCU
+ */
+static DEFINE_RWLOCK(pathtbl_resize_lock);
+
+/**
+ *
+ * mesh_path_assign_nexthop - update mesh path next hop
+ *
+ * @mpath: mesh path to update
+ * @sta: next hop to assign
+ *
+ * Locking: mpath->state_lock must be held when calling this function
+ */
+void mesh_path_assign_nexthop(struct mesh_path *mpath, struct sta_info *sta)
+{
+	__sta_info_get(sta);
+	if (mpath->next_hop)
+		sta_info_put(mpath->next_hop);
+	mpath->next_hop = sta;
+}
+
+
+/**
+ * mesh_path_lookup - look up a path in the mesh path table
+ * @dst: hardware address (ETH_ALEN length) of destination
+ * @dev: local interface
+ *
+ * Returns: pointer to the mesh path structure, or NULL if not found
+ *
+ * Locking: must be called within a read rcu section.
+ */
+struct mesh_path *mesh_path_lookup(u8 *dst, struct net_device *dev)
+{
+	struct mesh_path *mpath;
+	struct hlist_node *n;
+	struct hlist_head *bucket;
+	struct mesh_table *tbl;
+	struct mpath_node *node;
+
+	tbl = rcu_dereference(mesh_paths);
+
+	bucket = &tbl->hash_buckets[mesh_table_hash(dst, dev, tbl)];
+	hlist_for_each_entry_rcu(node, n, bucket, list) {
+		mpath = node->mpath;
+		if (mpath->dev == dev &&
+				memcmp(dst, mpath->dst, ETH_ALEN) == 0) {
+			if (MPATH_EXPIRED(mpath)) {
+				spin_lock_bh(&mpath->state_lock);
+				if (MPATH_EXPIRED(mpath))
+					mpath->flags &= ~MESH_PATH_ACTIVE;
+				spin_unlock_bh(&mpath->state_lock);
+			}
+			return mpath;
+		}
+	}
+	return NULL;
+}
+
+/**
+ * mesh_path_lookup_by_idx - look up a path in the mesh path table by its index
+ * @idx: index
+ * @dev: local interface
+ *
+ * Returns: pointer to the mesh path structure, or NULL if not found.
+ *
+ * Locking: must be called within a read rcu section.
+ */
+struct mesh_path *mesh_path_lookup_by_idx(int idx, struct net_device *dev)
+{
+	struct mpath_node *node;
+	struct hlist_node *p;
+	int i;
+	int j = 0;
+
+	for_each_mesh_entry(mesh_paths, p, node, i)
+		if (j++ == idx) {
+			if (MPATH_EXPIRED(node->mpath)) {
+				spin_lock_bh(&node->mpath->state_lock);
+				if (MPATH_EXPIRED(node->mpath))
+					node->mpath->flags &= ~MESH_PATH_ACTIVE;
+				spin_unlock_bh(&node->mpath->state_lock);
+			}
+			return node->mpath;
+		}
+
+	return NULL;
+}
+
+/**
+ * mesh_path_add - allocate and add a new path to the mesh path table
+ * @addr: destination address of the path (ETH_ALEN length)
+ * @dev: local interface
+ *
+ * Returns: 0 on sucess
+ *
+ * State: the initial state of the new path is set to 0
+ */
+int mesh_path_add(u8 *dst, struct net_device *dev)
+{
+	struct ieee80211_sub_if_data *sdata = IEEE80211_DEV_TO_SUB_IF(dev);
+	struct mesh_path *mpath, *new_mpath;
+	struct mpath_node *node, *new_node;
+	struct hlist_head *bucket;
+	struct hlist_node *n;
+	int grow = 0;
+	int err = 0;
+	u32 hash_idx;
+
+	if (memcmp(dst, dev->dev_addr, ETH_ALEN) == 0)
+		/* never add ourselves as neighbours */
+		return -ENOTSUPP;
+
+	if (is_multicast_ether_addr(dst))
+		return -ENOTSUPP;
+
+	if (atomic_add_unless(&sdata->u.sta.mpaths, 1, MESH_MAX_MPATHS) == 0)
+		return -ENOSPC;
+
+	read_lock(&pathtbl_resize_lock);
+
+	new_mpath = kzalloc(sizeof(struct mesh_path), GFP_KERNEL);
+	if (!new_mpath) {
+		atomic_dec(&sdata->u.sta.mpaths);
+		err = -ENOMEM;
+		goto endadd2;
+	}
+	memcpy(new_mpath->dst, dst, ETH_ALEN);
+	new_mpath->dev = dev;
+	new_mpath->flags = 0;
+	skb_queue_head_init(&new_mpath->frame_queue);
+	new_node = kmalloc(sizeof(struct mpath_node), GFP_KERNEL);
+	new_node->mpath = new_mpath;
+	new_mpath->timer.data = (unsigned long) new_mpath;
+	new_mpath->timer.function = mesh_path_timer;
+	new_mpath->exp_time = jiffies;
+	spin_lock_init(&new_mpath->state_lock);
+	init_timer(&new_mpath->timer);
+
+	hash_idx = mesh_table_hash(dst, dev, mesh_paths);
+	bucket = &mesh_paths->hash_buckets[hash_idx];
+
+	spin_lock(&mesh_paths->hashwlock[hash_idx]);
+
+	hlist_for_each_entry(node, n, bucket, list) {
+		mpath = node->mpath;
+		if (mpath->dev == dev && memcmp(dst, mpath->dst, ETH_ALEN)
+				== 0) {
+			err = -EEXIST;
+			atomic_dec(&sdata->u.sta.mpaths);
+			kfree(new_node);
+			kfree(new_mpath);
+			goto endadd;
+		}
+	}
+
+	hlist_add_head_rcu(&new_node->list, bucket);
+	if (atomic_inc_return(&mesh_paths->entries) >=
+		mesh_paths->mean_chain_len * (mesh_paths->hash_mask + 1))
+		grow = 1;
+
+endadd:
+	spin_unlock(&mesh_paths->hashwlock[hash_idx]);
+endadd2:
+	read_unlock(&pathtbl_resize_lock);
+	if (!err && grow) {
+		struct mesh_table *oldtbl, *newtbl;
+
+		write_lock(&pathtbl_resize_lock);
+		oldtbl = mesh_paths;
+		newtbl = mesh_table_grow(mesh_paths);
+		if (!newtbl) {
+			write_unlock(&pathtbl_resize_lock);
+			return -ENOMEM;
+		}
+		rcu_assign_pointer(mesh_paths, newtbl);
+		synchronize_rcu();
+		mesh_table_free(oldtbl, false);
+		write_unlock(&pathtbl_resize_lock);
+	}
+	return err;
+}
+
+
+/**
+ * mesh_plink_broken - deactivates paths and sends perr when a link breaks
+ *
+ * @sta: broken peer link
+ *
+ * This function must be called from the rate control algorithm if enough
+ * delivery errors suggest that a peer link is no longer usable.
+ */
+void mesh_plink_broken(struct sta_info *sta)
+{
+	struct mesh_path *mpath;
+	struct mpath_node *node;
+	struct hlist_node *p;
+	struct net_device *dev = sta->dev;
+	int i;
+
+	rcu_read_lock();
+	for_each_mesh_entry(mesh_paths, p, node, i) {
+		mpath = node->mpath;
+		spin_lock_bh(&mpath->state_lock);
+		if (mpath->next_hop == sta &&
+		    mpath->flags & MESH_PATH_ACTIVE &&
+		    !(mpath->flags & MESH_PATH_FIXED)) {
+			mpath->flags &= ~MESH_PATH_ACTIVE;
+			++mpath->dsn;
+			spin_unlock_bh(&mpath->state_lock);
+			mesh_path_error_tx(mpath->dst,
+					cpu_to_le32(mpath->dsn),
+					dev->broadcast, dev);
+		} else
+		spin_unlock_bh(&mpath->state_lock);
+	}
+	rcu_read_unlock();
+}
+
+/**
+ * mesh_path_flush_by_nexthop - Deletes mesh paths if their next hop matches
+ *
+ * @sta - mesh peer to match
+ *
+ * RCU notes: this function is called when a mesh plink transitions from ESTAB
+ * to any other state, since ESTAB state is the only one that allows path
+ * creation. This will happen before the sta can be freed (since we hold
+ * a reference to it) so any reader in a rcu read block will be protected
+ * against the plink dissapearing.
+ */
+void mesh_path_flush_by_nexthop(struct sta_info *sta)
+{
+	struct mesh_path *mpath;
+	struct mpath_node *node;
+	struct hlist_node *p;
+	int i;
+
+	for_each_mesh_entry(mesh_paths, p, node, i) {
+		mpath = node->mpath;
+		if (mpath->next_hop == sta)
+			mesh_path_del(mpath->dst, mpath->dev);
+	}
+}
+
+void mesh_path_flush(struct net_device *dev)
+{
+	struct mesh_path *mpath;
+	struct mpath_node *node;
+	struct hlist_node *p;
+	int i;
+
+	for_each_mesh_entry(mesh_paths, p, node, i) {
+		mpath = node->mpath;
+		if (mpath->dev == dev)
+			mesh_path_del(mpath->dst, mpath->dev);
+	}
+}
+
+static void mesh_path_node_reclaim(struct rcu_head *rp)
+{
+	struct mpath_node *node = container_of(rp, struct mpath_node, rcu);
+	struct ieee80211_sub_if_data *sdata =
+		IEEE80211_DEV_TO_SUB_IF(node->mpath->dev);
+	if (node->mpath->next_hop)
+		sta_info_put(node->mpath->next_hop);
+	atomic_dec(&sdata->u.sta.mpaths);
+	kfree(node->mpath);
+	kfree(node);
+}
+
+/**
+ * mesh_path_del - delete a mesh path from the table
+ *
+ * @addr: dst address (ETH_ALEN length)
+ * @dev: local interface
+ *
+ * Returns: 0 if succesful
+ *
+ * State: if the path is being resolved, the deletion will be postponed until
+ * the path resolution completes or times out.
+ */
+int mesh_path_del(u8 *addr, struct net_device *dev)
+{
+	struct mesh_path *mpath;
+	struct mpath_node *node;
+	struct hlist_head *bucket;
+	struct hlist_node *n;
+	int hash_idx;
+	int err = 0;
+
+	read_lock(&pathtbl_resize_lock);
+	hash_idx = mesh_table_hash(addr, dev, mesh_paths);
+	bucket = &mesh_paths->hash_buckets[hash_idx];
+
+	spin_lock(&mesh_paths->hashwlock[hash_idx]);
+	hlist_for_each_entry(node, n, bucket, list) {
+		mpath = node->mpath;
+		if (mpath->dev == dev &&
+				memcmp(addr, mpath->dst, ETH_ALEN) == 0) {
+			spin_lock_bh(&mpath->state_lock);
+			if (mpath->flags & MESH_PATH_RESOLVING) {
+				mpath->flags |= MESH_PATH_DELETE;
+			} else {
+				mpath->flags |= MESH_PATH_RESOLVING;
+				hlist_del_rcu(&node->list);
+				call_rcu(&node->rcu, mesh_path_node_reclaim);
+				atomic_dec(&mesh_paths->entries);
+			}
+			spin_unlock_bh(&mpath->state_lock);
+			goto enddel;
+		}
+	}
+
+	err = -ENXIO;
+enddel:
+	spin_unlock(&mesh_paths->hashwlock[hash_idx]);
+	read_unlock(&pathtbl_resize_lock);
+	return err;
+}
+
+/**
+ * mesh_path_tx_pending - sends pending frames in a mesh path queue
+ *
+ * @mpath: mesh path to activate
+ *
+ * Locking: the state_lock of the mpath structure must NOT be held when calling
+ * this function.
+ */
+void mesh_path_tx_pending(struct mesh_path *mpath)
+{
+	struct sk_buff *skb;
+
+	while ((skb = skb_dequeue(&mpath->frame_queue)) &&
+			(mpath->flags & MESH_PATH_ACTIVE))
+		dev_queue_xmit(skb);
+}
+
+/**
+ * mesh_path_discard_frame - discard a frame whose path could not be resolved
+ *
+ * @skb: frame to discard
+ * @dev: network device the frame was to be sent through
+ *
+ * If the frame was beign forwarded from another MP, a PERR frame will be sent
+ * to the precursor.
+ *
+ * Locking: the function must me called within a rcu_read_lock region
+ */
+void mesh_path_discard_frame(struct sk_buff *skb, struct net_device *dev)
+{
+	struct ieee80211_sub_if_data *sdata = IEEE80211_DEV_TO_SUB_IF(dev);
+	struct mesh_path *mpath;
+	u32 dsn = 0;
+
+	if (skb->pkt_type == PACKET_OTHERHOST) {
+		struct ieee80211s_hdr *prev_meshhdr;
+		int mshhdrlen;
+		u8 *ra, *da;
+
+		prev_meshhdr = ((struct ieee80211s_hdr *)skb->cb);
+		mshhdrlen = ieee80211_get_mesh_hdrlen(prev_meshhdr);
+		da = skb->data;
+		ra = MESH_PREQ(skb);
+		mpath = mesh_path_lookup(da, dev);
+		if (mpath)
+			dsn = ++mpath->dsn;
+		mesh_path_error_tx(skb->data, cpu_to_le32(dsn), ra, dev);
+	}
+
+	kfree_skb(skb);
+	sdata->u.sta.mshstats.dropped_frames_no_route++;
+}
+
+/**
+ * mesh_path_flush_pending - free the pending queue of a mesh path
+ *
+ * @mpath: mesh path whose queue has to be freed
+ *
+ * Locking: the function must me called withing a rcu_read_lock region
+ */
+void mesh_path_flush_pending(struct mesh_path *mpath)
+{
+	struct ieee80211_sub_if_data *sdata;
+	struct sk_buff *skb;
+
+	sdata = IEEE80211_DEV_TO_SUB_IF(mpath->dev);
+
+	while ((skb = skb_dequeue(&mpath->frame_queue)) &&
+			(mpath->flags & MESH_PATH_ACTIVE))
+		mesh_path_discard_frame(skb, mpath->dev);
+}
+
+/**
+ * mesh_path_fix_nexthop - force a specific next hop for a mesh path
+ *
+ * @mpath: the mesh path to modify
+ * @next_hop: the next hop to force
+ *
+ * Locking: this function must be called holding mpath->state_lock
+ */
+void mesh_path_fix_nexthop(struct mesh_path *mpath, struct sta_info *next_hop)
+{
+	spin_lock_bh(&mpath->state_lock);
+	mesh_path_assign_nexthop(mpath, next_hop);
+	mpath->dsn = 0xffff;
+	mpath->metric = 0;
+	mpath->hop_count = 0;
+	mpath->exp_time = 0;
+	mpath->flags |= MESH_PATH_FIXED;
+	mesh_path_activate(mpath);
+	spin_unlock_bh(&mpath->state_lock);
+	mesh_path_tx_pending(mpath);
+}
+
+static void mesh_path_node_free(struct hlist_node *p, bool free_leafs)
+{
+	struct mesh_path *mpath;
+	struct mpath_node *node = hlist_entry(p, struct mpath_node, list);
+	mpath = node->mpath;
+	hlist_del_rcu(p);
+	synchronize_rcu();
+	if (free_leafs)
+		kfree(mpath);
+	kfree(node);
+}
+
+static void mesh_path_node_copy(struct hlist_node *p, struct mesh_table *newtbl)
+{
+	struct mesh_path *mpath;
+	struct mpath_node *node, *new_node;
+	u32 hash_idx;
+
+	node = hlist_entry(p, struct mpath_node, list);
+	mpath = node->mpath;
+	new_node = kmalloc(sizeof(struct mpath_node), GFP_KERNEL);
+	new_node->mpath = mpath;
+	hash_idx = mesh_table_hash(mpath->dst, mpath->dev, newtbl);
+	hlist_add_head(&new_node->list,
+			&newtbl->hash_buckets[hash_idx]);
+}
+
+int mesh_pathtbl_init(void)
+{
+	mesh_paths = mesh_table_alloc(INIT_PATHS_SIZE_ORDER);
+	mesh_paths->free_node = &mesh_path_node_free;
+	mesh_paths->copy_node = &mesh_path_node_copy;
+	mesh_paths->mean_chain_len = MEAN_CHAIN_LEN;
+	if (!mesh_paths)
+		return -ENOMEM;
+	return 0;
+}
+
+void mesh_path_expire(struct net_device *dev)
+{
+	struct mesh_path *mpath;
+	struct mpath_node *node;
+	struct hlist_node *p;
+	int i;
+
+	read_lock(&pathtbl_resize_lock);
+	for_each_mesh_entry(mesh_paths, p, node, i) {
+		if (node->mpath->dev != dev)
+			continue;
+		mpath = node->mpath;
+		spin_lock_bh(&mpath->state_lock);
+		if ((!(mpath->flags & MESH_PATH_RESOLVING)) &&
+		    (!(mpath->flags & MESH_PATH_FIXED)) &&
+			time_after(jiffies,
+			 mpath->exp_time + MESH_PATH_EXPIRE)) {
+			spin_unlock_bh(&mpath->state_lock);
+			mesh_path_del(mpath->dst, mpath->dev);
+		} else
+			spin_unlock_bh(&mpath->state_lock);
+	}
+	read_unlock(&pathtbl_resize_lock);
+}
+
+void mesh_pathtbl_unregister(void)
+{
+	mesh_table_free(mesh_paths, true);
+}