datapath: compat: Backports nf_conncount

[ovs.git] / datapath / linux / compat / nf_conncount.c

/*
 * Backported from upstream commit 5c789e131cbb ("netfilter:
 * nf_conncount: Add list lock and gc worker, and RCU for init tree search")
 *
 * count the number of connections matching an arbitrary key.
 *
 * (C) 2017 Red Hat GmbH
 * Author: Florian Westphal <fw@strlen.de>
 *
 * split from xt_connlimit.c:
 *   (c) 2000 Gerd Knorr <kraxel@bytesex.org>
 *   Nov 2002: Martin Bene <martin.bene@icomedias.com>:
 *              only ignore TIME_WAIT or gone connections
 *   (C) CC Computer Consultants GmbH, 2007
 */
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/in.h>
#include <linux/in6.h>
#include <linux/ip.h>
#include <linux/ipv6.h>
#include <linux/jhash.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <linux/rbtree.h>
#include <linux/module.h>
#include <linux/random.h>
#include <linux/skbuff.h>
#include <linux/spinlock.h>
#include <linux/netfilter/nf_conntrack_tcp.h>
#include <linux/netfilter/x_tables.h>
#include <net/netfilter/nf_conntrack.h>
#include <net/netfilter/nf_conntrack_count.h>
#include <net/netfilter/nf_conntrack_core.h>
#include <net/netfilter/nf_conntrack_tuple.h>
#include <net/netfilter/nf_conntrack_zones.h>

#define CONNCOUNT_SLOTS         256U

#ifdef CONFIG_LOCKDEP
#define CONNCOUNT_LOCK_SLOTS    8U
#else
#define CONNCOUNT_LOCK_SLOTS    256U
#endif

#define CONNCOUNT_GC_MAX_NODES  8
#define MAX_KEYLEN              5

/* we will save the tuples of all connections we care about */
struct nf_conncount_tuple {
        struct list_head                node;
        struct nf_conntrack_tuple       tuple;
        struct nf_conntrack_zone        zone;
        int                             cpu;
        u32                             jiffies32;
        struct rcu_head                 rcu_head;
};

struct nf_conncount_rb {
        struct rb_node node;
        struct nf_conncount_list list;
        u32 key[MAX_KEYLEN];
        struct rcu_head rcu_head;
};

static spinlock_t nf_conncount_locks[CONNCOUNT_LOCK_SLOTS] __cacheline_aligned_in_smp;

struct nf_conncount_data {
        unsigned int keylen;
        struct rb_root root[CONNCOUNT_SLOTS];
        struct net *net;
        struct work_struct gc_work;
        unsigned long pending_trees[BITS_TO_LONGS(CONNCOUNT_SLOTS)];
        unsigned int gc_tree;
};

static u_int32_t conncount_rnd __read_mostly;
static struct kmem_cache *conncount_rb_cachep __read_mostly;
static struct kmem_cache *conncount_conn_cachep __read_mostly;

static inline bool already_closed(const struct nf_conn *conn)
{
        if (nf_ct_protonum(conn) == IPPROTO_TCP)
                return conn->proto.tcp.state == TCP_CONNTRACK_TIME_WAIT ||
                       conn->proto.tcp.state == TCP_CONNTRACK_CLOSE;
        else
                return false;
}

static int key_diff(const u32 *a, const u32 *b, unsigned int klen)
{
        return memcmp(a, b, klen * sizeof(u32));
}

static enum nf_conncount_list_add
nf_conncount_add(struct nf_conncount_list *list,
                 const struct nf_conntrack_tuple *tuple,
                 const struct nf_conntrack_zone *zone)
{
        struct nf_conncount_tuple *conn;

        if (WARN_ON_ONCE(list->count > INT_MAX))
                return NF_CONNCOUNT_ERR;

        conn = kmem_cache_alloc(conncount_conn_cachep, GFP_ATOMIC);
        if (conn == NULL)
                return NF_CONNCOUNT_ERR;

        conn->tuple = *tuple;
        conn->zone = *zone;
        conn->cpu = raw_smp_processor_id();
        conn->jiffies32 = (u32)jiffies;
        spin_lock(&list->list_lock);
        if (list->dead == true) {
                kmem_cache_free(conncount_conn_cachep, conn);
                spin_unlock(&list->list_lock);
                return NF_CONNCOUNT_SKIP;
        }
        list_add_tail(&conn->node, &list->head);
        list->count++;
        spin_unlock(&list->list_lock);
        return NF_CONNCOUNT_ADDED;
}

static void __conn_free(struct rcu_head *h)
{
        struct nf_conncount_tuple *conn;

        conn = container_of(h, struct nf_conncount_tuple, rcu_head);
        kmem_cache_free(conncount_conn_cachep, conn);
}

static bool conn_free(struct nf_conncount_list *list,
                      struct nf_conncount_tuple *conn)
{
        bool free_entry = false;

        spin_lock(&list->list_lock);

        if (list->count == 0) {
                spin_unlock(&list->list_lock);
                return free_entry;
        }

        list->count--;
        list_del_rcu(&conn->node);
        if (list->count == 0)
                free_entry = true;

        spin_unlock(&list->list_lock);
        call_rcu(&conn->rcu_head, __conn_free);
        return free_entry;
}

static const struct nf_conntrack_tuple_hash *
find_or_evict(struct net *net, struct nf_conncount_list *list,
              struct nf_conncount_tuple *conn, bool *free_entry)
{
        const struct nf_conntrack_tuple_hash *found;
        unsigned long a, b;
        int cpu = raw_smp_processor_id();
        __s32 age;

        found = nf_conntrack_find_get(net, &conn->zone, &conn->tuple);
        if (found)
                return found;
        b = conn->jiffies32;
        a = (u32)jiffies;

        /* conn might have been added just before by another cpu and
         * might still be unconfirmed.  In this case, nf_conntrack_find()
         * returns no result.  Thus only evict if this cpu added the
         * stale entry or if the entry is older than two jiffies.
         */
        age = a - b;
        if (conn->cpu == cpu || age >= 2) {
                *free_entry = conn_free(list, conn);
                return ERR_PTR(-ENOENT);
        }

        return ERR_PTR(-EAGAIN);
}

static void nf_conncount_lookup(struct net *net,
                         struct nf_conncount_list *list,
                         const struct nf_conntrack_tuple *tuple,
                         const struct nf_conntrack_zone *zone,
                         bool *addit)
{
        const struct nf_conntrack_tuple_hash *found;
        struct nf_conncount_tuple *conn, *conn_n;
        struct nf_conn *found_ct;
        unsigned int collect = 0;
        bool free_entry = false;

        /* best effort only */
        *addit = tuple ? true : false;

        /* check the saved connections */
        list_for_each_entry_safe(conn, conn_n, &list->head, node) {
                if (collect > CONNCOUNT_GC_MAX_NODES)
                        break;

                found = find_or_evict(net, list, conn, &free_entry);
                if (IS_ERR(found)) {
                        /* Not found, but might be about to be confirmed */
                        if (PTR_ERR(found) == -EAGAIN) {
                                if (!tuple)
                                        continue;

                                if (nf_ct_tuple_equal(&conn->tuple, tuple) &&
                                    nf_ct_zone_id(&conn->zone, conn->zone.dir) ==
                                    nf_ct_zone_id(zone, zone->dir))
                                        *addit = false;
                        } else if (PTR_ERR(found) == -ENOENT)
                                collect++;
                        continue;
                }

                found_ct = nf_ct_tuplehash_to_ctrack(found);

                if (tuple && nf_ct_tuple_equal(&conn->tuple, tuple) &&
                    nf_ct_zone_equal(found_ct, zone, zone->dir)) {
                        /*
                         * We should not see tuples twice unless someone hooks
                         * this into a table without "-p tcp --syn".
                         *
                         * Attempt to avoid a re-add in this case.
                         */
                        *addit = false;
                } else if (already_closed(found_ct)) {
                        /*
                         * we do not care about connections which are
                         * closed already -> ditch it
                         */
                        nf_ct_put(found_ct);
                        conn_free(list, conn);
                        collect++;
                        continue;
                }

                nf_ct_put(found_ct);
        }
}

static void nf_conncount_list_init(struct nf_conncount_list *list)
{
        spin_lock_init(&list->list_lock);
        INIT_LIST_HEAD(&list->head);
        list->count = 1;
        list->dead = false;
}

/* Return true if the list is empty */
static bool nf_conncount_gc_list(struct net *net,
                          struct nf_conncount_list *list)
{
        const struct nf_conntrack_tuple_hash *found;
        struct nf_conncount_tuple *conn, *conn_n;
        struct nf_conn *found_ct;
        unsigned int collected = 0;
        bool free_entry = false;

        list_for_each_entry_safe(conn, conn_n, &list->head, node) {
                found = find_or_evict(net, list, conn, &free_entry);
                if (IS_ERR(found)) {
                        if (PTR_ERR(found) == -ENOENT)  {
                                if (free_entry)
                                        return true;
                                collected++;
                        }
                        continue;
                }

                found_ct = nf_ct_tuplehash_to_ctrack(found);
                if (already_closed(found_ct)) {
                        /*
                         * we do not care about connections which are
                         * closed already -> ditch it
                         */
                        nf_ct_put(found_ct);
                        if (conn_free(list, conn))
                                return true;
                        collected++;
                        continue;
                }

                nf_ct_put(found_ct);
                if (collected > CONNCOUNT_GC_MAX_NODES)
                        return false;
        }
        return false;
}

static void __tree_nodes_free(struct rcu_head *h)
{
        struct nf_conncount_rb *rbconn;

        rbconn = container_of(h, struct nf_conncount_rb, rcu_head);
        kmem_cache_free(conncount_rb_cachep, rbconn);
}

static void tree_nodes_free(struct rb_root *root,
                            struct nf_conncount_rb *gc_nodes[],
                            unsigned int gc_count)
{
        struct nf_conncount_rb *rbconn;

        while (gc_count) {
                rbconn = gc_nodes[--gc_count];
                spin_lock(&rbconn->list.list_lock);
                if (rbconn->list.count == 0 && rbconn->list.dead == false) {
                        rbconn->list.dead = true;
                        rb_erase(&rbconn->node, root);
                        call_rcu(&rbconn->rcu_head, __tree_nodes_free);
                }
                spin_unlock(&rbconn->list.list_lock);
        }
}

static void schedule_gc_worker(struct nf_conncount_data *data, int tree)
{
        set_bit(tree, data->pending_trees);
        schedule_work(&data->gc_work);
}

static unsigned int
insert_tree(struct net *net,
            struct nf_conncount_data *data,
            struct rb_root *root,
            unsigned int hash,
            const u32 *key,
            u8 keylen,
            const struct nf_conntrack_tuple *tuple,
            const struct nf_conntrack_zone *zone)
{
        enum nf_conncount_list_add ret;
        struct nf_conncount_rb *gc_nodes[CONNCOUNT_GC_MAX_NODES];
        struct rb_node **rbnode, *parent;
        struct nf_conncount_rb *rbconn;
        struct nf_conncount_tuple *conn;
        unsigned int count = 0, gc_count = 0;
        bool node_found = false;

        spin_lock_bh(&nf_conncount_locks[hash % CONNCOUNT_LOCK_SLOTS]);

        parent = NULL;
        rbnode = &(root->rb_node);
        while (*rbnode) {
                int diff;
                rbconn = rb_entry(*rbnode, struct nf_conncount_rb, node);

                parent = *rbnode;
                diff = key_diff(key, rbconn->key, keylen);
                if (diff < 0) {
                        rbnode = &((*rbnode)->rb_left);
                } else if (diff > 0) {
                        rbnode = &((*rbnode)->rb_right);
                } else {
                        /* unlikely: other cpu added node already */
                        node_found = true;
                        ret = nf_conncount_add(&rbconn->list, tuple, zone);
                        if (ret == NF_CONNCOUNT_ERR) {
                                count = 0; /* hotdrop */
                        } else if (ret == NF_CONNCOUNT_ADDED) {
                                count = rbconn->list.count;
                        } else {
                                /* NF_CONNCOUNT_SKIP, rbconn is already
                                 * reclaimed by gc, insert a new tree node
                                 */
                                node_found = false;
                        }
                        break;
                }

                if (gc_count >= ARRAY_SIZE(gc_nodes))
                        continue;

                if (nf_conncount_gc_list(net, &rbconn->list))
                        gc_nodes[gc_count++] = rbconn;
        }

        if (gc_count) {
                tree_nodes_free(root, gc_nodes, gc_count);
                /* tree_node_free before new allocation permits
                 * allocator to re-use newly free'd object.
                 *
                 * This is a rare event; in most cases we will find
                 * existing node to re-use. (or gc_count is 0).
                 */

                if (gc_count >= ARRAY_SIZE(gc_nodes))
                        schedule_gc_worker(data, hash);
        }

        if (node_found)
                goto out_unlock;

        /* expected case: match, insert new node */
        rbconn = kmem_cache_alloc(conncount_rb_cachep, GFP_ATOMIC);
        if (rbconn == NULL)
                goto out_unlock;

        conn = kmem_cache_alloc(conncount_conn_cachep, GFP_ATOMIC);
        if (conn == NULL) {
                kmem_cache_free(conncount_rb_cachep, rbconn);
                goto out_unlock;
        }

        conn->tuple = *tuple;
        conn->zone = *zone;
        memcpy(rbconn->key, key, sizeof(u32) * keylen);

        nf_conncount_list_init(&rbconn->list);
        list_add(&conn->node, &rbconn->list.head);
        count = 1;

        rb_link_node(&rbconn->node, parent, rbnode);
        rb_insert_color(&rbconn->node, root);
out_unlock:
        spin_unlock_bh(&nf_conncount_locks[hash % CONNCOUNT_LOCK_SLOTS]);
        return count;
}

static unsigned int
count_tree(struct net *net,
           struct nf_conncount_data *data,
           const u32 *key,
           const struct nf_conntrack_tuple *tuple,
           const struct nf_conntrack_zone *zone)
{
        enum nf_conncount_list_add ret;
        struct rb_root *root;
        struct rb_node *parent;
        struct nf_conncount_rb *rbconn;
        unsigned int hash;
        u8 keylen = data->keylen;

        hash = jhash2(key, data->keylen, conncount_rnd) % CONNCOUNT_SLOTS;
        root = &data->root[hash];

        parent = rcu_dereference_raw(root->rb_node);
        while (parent) {
                int diff;
                bool addit;

                rbconn = rb_entry(parent, struct nf_conncount_rb, node);

                diff = key_diff(key, rbconn->key, keylen);
                if (diff < 0) {
                        parent = rcu_dereference_raw(parent->rb_left);
                } else if (diff > 0) {
                        parent = rcu_dereference_raw(parent->rb_right);
                } else {
                        /* same source network -> be counted! */
                        nf_conncount_lookup(net, &rbconn->list, tuple, zone,
                                            &addit);

                        if (!addit)
                                return rbconn->list.count;

                        ret = nf_conncount_add(&rbconn->list, tuple, zone);
                        if (ret == NF_CONNCOUNT_ERR) {
                                return 0; /* hotdrop */
                        } else if (ret == NF_CONNCOUNT_ADDED) {
                                return rbconn->list.count;
                        } else {
                                /* NF_CONNCOUNT_SKIP, rbconn is already
                                 * reclaimed by gc, insert a new tree node
                                 */
                                break;
                        }
                }
        }

        if (!tuple)
                return 0;

        return insert_tree(net, data, root, hash, key, keylen, tuple, zone);
}

static void tree_gc_worker(struct work_struct *work)
{
        struct nf_conncount_data *data = container_of(work, struct nf_conncount_data, gc_work);
        struct nf_conncount_rb *gc_nodes[CONNCOUNT_GC_MAX_NODES], *rbconn;
        struct rb_root *root;
        struct rb_node *node;
        unsigned int tree, next_tree, gc_count = 0;

        tree = data->gc_tree % CONNCOUNT_LOCK_SLOTS;
        root = &data->root[tree];

        rcu_read_lock();
        for (node = rb_first(root); node != NULL; node = rb_next(node)) {
                rbconn = rb_entry(node, struct nf_conncount_rb, node);
                if (nf_conncount_gc_list(data->net, &rbconn->list))
                        gc_nodes[gc_count++] = rbconn;
        }
        rcu_read_unlock();

        spin_lock_bh(&nf_conncount_locks[tree]);

        if (gc_count) {
                tree_nodes_free(root, gc_nodes, gc_count);
        }

        clear_bit(tree, data->pending_trees);

        next_tree = (tree + 1) % CONNCOUNT_SLOTS;
        next_tree = find_next_bit(data->pending_trees, next_tree, CONNCOUNT_SLOTS);

        if (next_tree < CONNCOUNT_SLOTS) {
                data->gc_tree = next_tree;
                schedule_work(work);
        }

        spin_unlock_bh(&nf_conncount_locks[tree]);
}

/* Count and return number of conntrack entries in 'net' with particular 'key'.
 * If 'tuple' is not null, insert it into the accounting data structure.
 * Call with RCU read lock.
 */
unsigned int rpl_nf_conncount_count(struct net *net,
                                    struct nf_conncount_data *data,
                                    const u32 *key,
                                    const struct nf_conntrack_tuple *tuple,
                                    const struct nf_conntrack_zone *zone)
{
        return count_tree(net, data, key, tuple, zone);
}
EXPORT_SYMBOL_GPL(rpl_nf_conncount_count);

struct nf_conncount_data *rpl_nf_conncount_init(struct net *net, unsigned int family,
                                            unsigned int keylen)
{
        struct nf_conncount_data *data;
        int ret, i;

        if (keylen % sizeof(u32) ||
            keylen / sizeof(u32) > MAX_KEYLEN ||
            keylen == 0)
                return ERR_PTR(-EINVAL);

        net_get_random_once(&conncount_rnd, sizeof(conncount_rnd));

        data = kmalloc(sizeof(*data), GFP_KERNEL);
        if (!data)
                return ERR_PTR(-ENOMEM);

        ret = nf_ct_netns_get(net, family);
        if (ret < 0) {
                kfree(data);
                return ERR_PTR(ret);
        }

        for (i = 0; i < ARRAY_SIZE(data->root); ++i)
                data->root[i] = RB_ROOT;

        data->keylen = keylen / sizeof(u32);
        data->net = net;
        INIT_WORK(&data->gc_work, tree_gc_worker);

        return data;
}
EXPORT_SYMBOL_GPL(rpl_nf_conncount_init);

static void nf_conncount_cache_free(struct nf_conncount_list *list)
{
        struct nf_conncount_tuple *conn, *conn_n;

        list_for_each_entry_safe(conn, conn_n, &list->head, node)
                kmem_cache_free(conncount_conn_cachep, conn);
}

static void destroy_tree(struct rb_root *r)
{
        struct nf_conncount_rb *rbconn;
        struct rb_node *node;

        while ((node = rb_first(r)) != NULL) {
                rbconn = rb_entry(node, struct nf_conncount_rb, node);

                rb_erase(node, r);

                nf_conncount_cache_free(&rbconn->list);

                kmem_cache_free(conncount_rb_cachep, rbconn);
        }
}

void rpl_nf_conncount_destroy(struct net *net, unsigned int family,
                              struct nf_conncount_data *data)
{
        unsigned int i;

        cancel_work_sync(&data->gc_work);
        nf_ct_netns_put(net, family);

        for (i = 0; i < ARRAY_SIZE(data->root); ++i)
                destroy_tree(&data->root[i]);

        kfree(data);
}
EXPORT_SYMBOL_GPL(rpl_nf_conncount_destroy);

int rpl_nf_conncount_modinit(void)
{
        int i;

        BUILD_BUG_ON(CONNCOUNT_LOCK_SLOTS > CONNCOUNT_SLOTS);
        BUILD_BUG_ON((CONNCOUNT_SLOTS % CONNCOUNT_LOCK_SLOTS) != 0);

        for (i = 0; i < CONNCOUNT_LOCK_SLOTS; ++i)
                spin_lock_init(&nf_conncount_locks[i]);

        conncount_conn_cachep = kmem_cache_create("nf_conncount_tuple",
                                           sizeof(struct nf_conncount_tuple),
                                           0, 0, NULL);
        if (!conncount_conn_cachep)
                return -ENOMEM;

        conncount_rb_cachep = kmem_cache_create("nf_conncount_rb",
                                           sizeof(struct nf_conncount_rb),
                                           0, 0, NULL);
        if (!conncount_rb_cachep) {
                kmem_cache_destroy(conncount_conn_cachep);
                return -ENOMEM;
        }

        return 0;
}

void rpl_nf_conncount_modexit(void)
{
        kmem_cache_destroy(conncount_conn_cachep);
        kmem_cache_destroy(conncount_rb_cachep);
}