net/openvswitch/flow_table.c

0001 // SPDX-License-Identifier: GPL-2.0-only
0002 /*
0003  * Copyright (c) 2007-2014 Nicira, Inc.
0004  */
0005
0006 #include "flow.h"
0007 #include "datapath.h"
0008 #include "flow_netlink.h"
0009 #include <linux/uaccess.h>
0010 #include <linux/netdevice.h>
0011 #include <linux/etherdevice.h>
0012 #include <linux/if_ether.h>
0013 #include <linux/if_vlan.h>
0014 #include <net/llc_pdu.h>
0015 #include <linux/kernel.h>
0016 #include <linux/jhash.h>
0017 #include <linux/jiffies.h>
0018 #include <linux/llc.h>
0019 #include <linux/module.h>
0020 #include <linux/in.h>
0021 #include <linux/rcupdate.h>
0022 #include <linux/cpumask.h>
0023 #include <linux/if_arp.h>
0024 #include <linux/ip.h>
0025 #include <linux/ipv6.h>
0026 #include <linux/sctp.h>
0027 #include <linux/tcp.h>
0028 #include <linux/udp.h>
0029 #include <linux/icmp.h>
0030 #include <linux/icmpv6.h>
0031 #include <linux/rculist.h>
0032 #include <linux/sort.h>
0033 #include <net/ip.h>
0034 #include <net/ipv6.h>
0035 #include <net/ndisc.h>
0036
0037 #define TBL_MIN_BUCKETS     1024
0038 #define MASK_ARRAY_SIZE_MIN 16
0039 #define REHASH_INTERVAL     (10 * 60 * HZ)
0040
0041 #define MC_DEFAULT_HASH_ENTRIES 256
0042 #define MC_HASH_SHIFT       8
0043 #define MC_HASH_SEGS        ((sizeof(uint32_t) * 8) / MC_HASH_SHIFT)
0044
0045 static struct kmem_cache *flow_cache;
0046 struct kmem_cache *flow_stats_cache __read_mostly;
0047
0048 static u16 range_n_bytes(const struct sw_flow_key_range *range)
0049 {
0050     return range->end - range->start;
0051 }
0052
0053 void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
0054                bool full, const struct sw_flow_mask *mask)
0055 {
0056     int start = full ? 0 : mask->range.start;
0057     int len = full ? sizeof *dst : range_n_bytes(&mask->range);
0058     const long *m = (const long *)((const u8 *)&mask->key + start);
0059     const long *s = (const long *)((const u8 *)src + start);
0060     long *d = (long *)((u8 *)dst + start);
0061     int i;
0062
0063     /* If 'full' is true then all of 'dst' is fully initialized. Otherwise,
0064      * if 'full' is false the memory outside of the 'mask->range' is left
0065      * uninitialized. This can be used as an optimization when further
0066      * operations on 'dst' only use contents within 'mask->range'.
0067      */
0068     for (i = 0; i < len; i += sizeof(long))
0069         *d++ = *s++ & *m++;
0070 }
0071
0072 struct sw_flow *ovs_flow_alloc(void)
0073 {
0074     struct sw_flow *flow;
0075     struct sw_flow_stats *stats;
0076
0077     flow = kmem_cache_zalloc(flow_cache, GFP_KERNEL);
0078     if (!flow)
0079         return ERR_PTR(-ENOMEM);
0080
0081     flow->stats_last_writer = -1;
0082
0083     /* Initialize the default stat node. */
0084     stats = kmem_cache_alloc_node(flow_stats_cache,
0085                       GFP_KERNEL | __GFP_ZERO,
0086                       node_online(0) ? 0 : NUMA_NO_NODE);
0087     if (!stats)
0088         goto err;
0089
0090     spin_lock_init(&stats->lock);
0091
0092     RCU_INIT_POINTER(flow->stats[0], stats);
0093
0094     cpumask_set_cpu(0, &flow->cpu_used_mask);
0095
0096     return flow;
0097 err:
0098     kmem_cache_free(flow_cache, flow);
0099     return ERR_PTR(-ENOMEM);
0100 }
0101
0102 int ovs_flow_tbl_count(const struct flow_table *table)
0103 {
0104     return table->count;
0105 }
0106
0107 static void flow_free(struct sw_flow *flow)
0108 {
0109     int cpu;
0110
0111     if (ovs_identifier_is_key(&flow->id))
0112         kfree(flow->id.unmasked_key);
0113     if (flow->sf_acts)
0114         ovs_nla_free_flow_actions((struct sw_flow_actions __force *)
0115                       flow->sf_acts);
0116     /* We open code this to make sure cpu 0 is always considered */
0117     for (cpu = 0; cpu < nr_cpu_ids;
0118          cpu = cpumask_next(cpu, &flow->cpu_used_mask)) {
0119         if (flow->stats[cpu])
0120             kmem_cache_free(flow_stats_cache,
0121                     (struct sw_flow_stats __force *)flow->stats[cpu]);
0122     }
0123
0124     kmem_cache_free(flow_cache, flow);
0125 }
0126
0127 static void rcu_free_flow_callback(struct rcu_head *rcu)
0128 {
0129     struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
0130
0131     flow_free(flow);
0132 }
0133
0134 void ovs_flow_free(struct sw_flow *flow, bool deferred)
0135 {
0136     if (!flow)
0137         return;
0138
0139     if (deferred)
0140         call_rcu(&flow->rcu, rcu_free_flow_callback);
0141     else
0142         flow_free(flow);
0143 }
0144
0145 static void __table_instance_destroy(struct table_instance *ti)
0146 {
0147     kvfree(ti->buckets);
0148     kfree(ti);
0149 }
0150
0151 static struct table_instance *table_instance_alloc(int new_size)
0152 {
0153     struct table_instance *ti = kmalloc(sizeof(*ti), GFP_KERNEL);
0154     int i;
0155
0156     if (!ti)
0157         return NULL;
0158
0159     ti->buckets = kvmalloc_array(new_size, sizeof(struct hlist_head),
0160                      GFP_KERNEL);
0161     if (!ti->buckets) {
0162         kfree(ti);
0163         return NULL;
0164     }
0165
0166     for (i = 0; i < new_size; i++)
0167         INIT_HLIST_HEAD(&ti->buckets[i]);
0168
0169     ti->n_buckets = new_size;
0170     ti->node_ver = 0;
0171     get_random_bytes(&ti->hash_seed, sizeof(u32));
0172
0173     return ti;
0174 }
0175
0176 static void __mask_array_destroy(struct mask_array *ma)
0177 {
0178     free_percpu(ma->masks_usage_stats);
0179     kfree(ma);
0180 }
0181
0182 static void mask_array_rcu_cb(struct rcu_head *rcu)
0183 {
0184     struct mask_array *ma = container_of(rcu, struct mask_array, rcu);
0185
0186     __mask_array_destroy(ma);
0187 }
0188
0189 static void tbl_mask_array_reset_counters(struct mask_array *ma)
0190 {
0191     int i, cpu;
0192
0193     /* As the per CPU counters are not atomic we can not go ahead and
0194      * reset them from another CPU. To be able to still have an approximate
0195      * zero based counter we store the value at reset, and subtract it
0196      * later when processing.
0197      */
0198     for (i = 0; i < ma->max; i++) {
0199         ma->masks_usage_zero_cntr[i] = 0;
0200
0201         for_each_possible_cpu(cpu) {
0202             struct mask_array_stats *stats;
0203             unsigned int start;
0204             u64 counter;
0205
0206             stats = per_cpu_ptr(ma->masks_usage_stats, cpu);
0207             do {
0208                 start = u64_stats_fetch_begin_irq(&stats->syncp);
0209                 counter = stats->usage_cntrs[i];
0210             } while (u64_stats_fetch_retry_irq(&stats->syncp, start));
0211
0212             ma->masks_usage_zero_cntr[i] += counter;
0213         }
0214     }
0215 }
0216
0217 static struct mask_array *tbl_mask_array_alloc(int size)
0218 {
0219     struct mask_array *new;
0220
0221     size = max(MASK_ARRAY_SIZE_MIN, size);
0222     new = kzalloc(sizeof(struct mask_array) +
0223               sizeof(struct sw_flow_mask *) * size +
0224               sizeof(u64) * size, GFP_KERNEL);
0225     if (!new)
0226         return NULL;
0227
0228     new->masks_usage_zero_cntr = (u64 *)((u8 *)new +
0229                          sizeof(struct mask_array) +
0230                          sizeof(struct sw_flow_mask *) *
0231                          size);
0232
0233     new->masks_usage_stats = __alloc_percpu(sizeof(struct mask_array_stats) +
0234                         sizeof(u64) * size,
0235                         __alignof__(u64));
0236     if (!new->masks_usage_stats) {
0237         kfree(new);
0238         return NULL;
0239     }
0240
0241     new->count = 0;
0242     new->max = size;
0243
0244     return new;
0245 }
0246
0247 static int tbl_mask_array_realloc(struct flow_table *tbl, int size)
0248 {
0249     struct mask_array *old;
0250     struct mask_array *new;
0251
0252     new = tbl_mask_array_alloc(size);
0253     if (!new)
0254         return -ENOMEM;
0255
0256     old = ovsl_dereference(tbl->mask_array);
0257     if (old) {
0258         int i;
0259
0260         for (i = 0; i < old->max; i++) {
0261             if (ovsl_dereference(old->masks[i]))
0262                 new->masks[new->count++] = old->masks[i];
0263         }
0264         call_rcu(&old->rcu, mask_array_rcu_cb);
0265     }
0266
0267     rcu_assign_pointer(tbl->mask_array, new);
0268
0269     return 0;
0270 }
0271
0272 static int tbl_mask_array_add_mask(struct flow_table *tbl,
0273                    struct sw_flow_mask *new)
0274 {
0275     struct mask_array *ma = ovsl_dereference(tbl->mask_array);
0276     int err, ma_count = READ_ONCE(ma->count);
0277
0278     if (ma_count >= ma->max) {
0279         err = tbl_mask_array_realloc(tbl, ma->max +
0280                           MASK_ARRAY_SIZE_MIN);
0281         if (err)
0282             return err;
0283
0284         ma = ovsl_dereference(tbl->mask_array);
0285     } else {
0286         /* On every add or delete we need to reset the counters so
0287          * every new mask gets a fair chance of being prioritized.
0288          */
0289         tbl_mask_array_reset_counters(ma);
0290     }
0291
0292     BUG_ON(ovsl_dereference(ma->masks[ma_count]));
0293
0294     rcu_assign_pointer(ma->masks[ma_count], new);
0295     WRITE_ONCE(ma->count, ma_count + 1);
0296
0297     return 0;
0298 }
0299
0300 static void tbl_mask_array_del_mask(struct flow_table *tbl,
0301                     struct sw_flow_mask *mask)
0302 {
0303     struct mask_array *ma = ovsl_dereference(tbl->mask_array);
0304     int i, ma_count = READ_ONCE(ma->count);
0305
0306     /* Remove the deleted mask pointers from the array */
0307     for (i = 0; i < ma_count; i++) {
0308         if (mask == ovsl_dereference(ma->masks[i]))
0309             goto found;
0310     }
0311
0312     BUG();
0313     return;
0314
0315 found:
0316     WRITE_ONCE(ma->count, ma_count - 1);
0317
0318     rcu_assign_pointer(ma->masks[i], ma->masks[ma_count - 1]);
0319     RCU_INIT_POINTER(ma->masks[ma_count - 1], NULL);
0320
0321     kfree_rcu(mask, rcu);
0322
0323     /* Shrink the mask array if necessary. */
0324     if (ma->max >= (MASK_ARRAY_SIZE_MIN * 2) &&
0325         ma_count <= (ma->max / 3))
0326         tbl_mask_array_realloc(tbl, ma->max / 2);
0327     else
0328         tbl_mask_array_reset_counters(ma);
0329
0330 }
0331
0332 /* Remove 'mask' from the mask list, if it is not needed any more. */
0333 static void flow_mask_remove(struct flow_table *tbl, struct sw_flow_mask *mask)
0334 {
0335     if (mask) {
0336         /* ovs-lock is required to protect mask-refcount and
0337          * mask list.
0338          */
0339         ASSERT_OVSL();
0340         BUG_ON(!mask->ref_count);
0341         mask->ref_count--;
0342
0343         if (!mask->ref_count)
0344             tbl_mask_array_del_mask(tbl, mask);
0345     }
0346 }
0347
0348 static void __mask_cache_destroy(struct mask_cache *mc)
0349 {
0350     free_percpu(mc->mask_cache);
0351     kfree(mc);
0352 }
0353
0354 static void mask_cache_rcu_cb(struct rcu_head *rcu)
0355 {
0356     struct mask_cache *mc = container_of(rcu, struct mask_cache, rcu);
0357
0358     __mask_cache_destroy(mc);
0359 }
0360
0361 static struct mask_cache *tbl_mask_cache_alloc(u32 size)
0362 {
0363     struct mask_cache_entry __percpu *cache = NULL;
0364     struct mask_cache *new;
0365
0366     /* Only allow size to be 0, or a power of 2, and does not exceed
0367      * percpu allocation size.
0368      */
0369     if ((!is_power_of_2(size) && size != 0) ||
0370         (size * sizeof(struct mask_cache_entry)) > PCPU_MIN_UNIT_SIZE)
0371         return NULL;
0372
0373     new = kzalloc(sizeof(*new), GFP_KERNEL);
0374     if (!new)
0375         return NULL;
0376
0377     new->cache_size = size;
0378     if (new->cache_size > 0) {
0379         cache = __alloc_percpu(array_size(sizeof(struct mask_cache_entry),
0380                           new->cache_size),
0381                        __alignof__(struct mask_cache_entry));
0382         if (!cache) {
0383             kfree(new);
0384             return NULL;
0385         }
0386     }
0387
0388     new->mask_cache = cache;
0389     return new;
0390 }
0391 int ovs_flow_tbl_masks_cache_resize(struct flow_table *table, u32 size)
0392 {
0393     struct mask_cache *mc = rcu_dereference_ovsl(table->mask_cache);
0394     struct mask_cache *new;
0395
0396     if (size == mc->cache_size)
0397         return 0;
0398
0399     if ((!is_power_of_2(size) && size != 0) ||
0400         (size * sizeof(struct mask_cache_entry)) > PCPU_MIN_UNIT_SIZE)
0401         return -EINVAL;
0402
0403     new = tbl_mask_cache_alloc(size);
0404     if (!new)
0405         return -ENOMEM;
0406
0407     rcu_assign_pointer(table->mask_cache, new);
0408     call_rcu(&mc->rcu, mask_cache_rcu_cb);
0409
0410     return 0;
0411 }
0412
0413 int ovs_flow_tbl_init(struct flow_table *table)
0414 {
0415     struct table_instance *ti, *ufid_ti;
0416     struct mask_cache *mc;
0417     struct mask_array *ma;
0418
0419     mc = tbl_mask_cache_alloc(MC_DEFAULT_HASH_ENTRIES);
0420     if (!mc)
0421         return -ENOMEM;
0422
0423     ma = tbl_mask_array_alloc(MASK_ARRAY_SIZE_MIN);
0424     if (!ma)
0425         goto free_mask_cache;
0426
0427     ti = table_instance_alloc(TBL_MIN_BUCKETS);
0428     if (!ti)
0429         goto free_mask_array;
0430
0431     ufid_ti = table_instance_alloc(TBL_MIN_BUCKETS);
0432     if (!ufid_ti)
0433         goto free_ti;
0434
0435     rcu_assign_pointer(table->ti, ti);
0436     rcu_assign_pointer(table->ufid_ti, ufid_ti);
0437     rcu_assign_pointer(table->mask_array, ma);
0438     rcu_assign_pointer(table->mask_cache, mc);
0439     table->last_rehash = jiffies;
0440     table->count = 0;
0441     table->ufid_count = 0;
0442     return 0;
0443
0444 free_ti:
0445     __table_instance_destroy(ti);
0446 free_mask_array:
0447     __mask_array_destroy(ma);
0448 free_mask_cache:
0449     __mask_cache_destroy(mc);
0450     return -ENOMEM;
0451 }
0452
0453 static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
0454 {
0455     struct table_instance *ti;
0456
0457     ti = container_of(rcu, struct table_instance, rcu);
0458     __table_instance_destroy(ti);
0459 }
0460
0461 static void table_instance_flow_free(struct flow_table *table,
0462                      struct table_instance *ti,
0463                      struct table_instance *ufid_ti,
0464                      struct sw_flow *flow)
0465 {
0466     hlist_del_rcu(&flow->flow_table.node[ti->node_ver]);
0467     table->count--;
0468
0469     if (ovs_identifier_is_ufid(&flow->id)) {
0470         hlist_del_rcu(&flow->ufid_table.node[ufid_ti->node_ver]);
0471         table->ufid_count--;
0472     }
0473
0474     flow_mask_remove(table, flow->mask);
0475 }
0476
0477 /* Must be called with OVS mutex held. */
0478 void table_instance_flow_flush(struct flow_table *table,
0479                    struct table_instance *ti,
0480                    struct table_instance *ufid_ti)
0481 {
0482     int i;
0483
0484     for (i = 0; i < ti->n_buckets; i++) {
0485         struct hlist_head *head = &ti->buckets[i];
0486         struct hlist_node *n;
0487         struct sw_flow *flow;
0488
0489         hlist_for_each_entry_safe(flow, n, head,
0490                       flow_table.node[ti->node_ver]) {
0491
0492             table_instance_flow_free(table, ti, ufid_ti,
0493                          flow);
0494             ovs_flow_free(flow, true);
0495         }
0496     }
0497
0498     if (WARN_ON(table->count != 0 ||
0499             table->ufid_count != 0)) {
0500         table->count = 0;
0501         table->ufid_count = 0;
0502     }
0503 }
0504
0505 static void table_instance_destroy(struct table_instance *ti,
0506                    struct table_instance *ufid_ti)
0507 {
0508     call_rcu(&ti->rcu, flow_tbl_destroy_rcu_cb);
0509     call_rcu(&ufid_ti->rcu, flow_tbl_destroy_rcu_cb);
0510 }
0511
0512 /* No need for locking this function is called from RCU callback or
0513  * error path.
0514  */
0515 void ovs_flow_tbl_destroy(struct flow_table *table)
0516 {
0517     struct table_instance *ti = rcu_dereference_raw(table->ti);
0518     struct table_instance *ufid_ti = rcu_dereference_raw(table->ufid_ti);
0519     struct mask_cache *mc = rcu_dereference_raw(table->mask_cache);
0520     struct mask_array *ma = rcu_dereference_raw(table->mask_array);
0521
0522     call_rcu(&mc->rcu, mask_cache_rcu_cb);
0523     call_rcu(&ma->rcu, mask_array_rcu_cb);
0524     table_instance_destroy(ti, ufid_ti);
0525 }
0526
0527 struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *ti,
0528                        u32 *bucket, u32 *last)
0529 {
0530     struct sw_flow *flow;
0531     struct hlist_head *head;
0532     int ver;
0533     int i;
0534
0535     ver = ti->node_ver;
0536     while (*bucket < ti->n_buckets) {
0537         i = 0;
0538         head = &ti->buckets[*bucket];
0539         hlist_for_each_entry_rcu(flow, head, flow_table.node[ver]) {
0540             if (i < *last) {
0541                 i++;
0542                 continue;
0543             }
0544             *last = i + 1;
0545             return flow;
0546         }
0547         (*bucket)++;
0548         *last = 0;
0549     }
0550
0551     return NULL;
0552 }
0553
0554 static struct hlist_head *find_bucket(struct table_instance *ti, u32 hash)
0555 {
0556     hash = jhash_1word(hash, ti->hash_seed);
0557     return &ti->buckets[hash & (ti->n_buckets - 1)];
0558 }
0559
0560 static void table_instance_insert(struct table_instance *ti,
0561                   struct sw_flow *flow)
0562 {
0563     struct hlist_head *head;
0564
0565     head = find_bucket(ti, flow->flow_table.hash);
0566     hlist_add_head_rcu(&flow->flow_table.node[ti->node_ver], head);
0567 }
0568
0569 static void ufid_table_instance_insert(struct table_instance *ti,
0570                        struct sw_flow *flow)
0571 {
0572     struct hlist_head *head;
0573
0574     head = find_bucket(ti, flow->ufid_table.hash);
0575     hlist_add_head_rcu(&flow->ufid_table.node[ti->node_ver], head);
0576 }
0577
0578 static void flow_table_copy_flows(struct table_instance *old,
0579                   struct table_instance *new, bool ufid)
0580 {
0581     int old_ver;
0582     int i;
0583
0584     old_ver = old->node_ver;
0585     new->node_ver = !old_ver;
0586
0587     /* Insert in new table. */
0588     for (i = 0; i < old->n_buckets; i++) {
0589         struct sw_flow *flow;
0590         struct hlist_head *head = &old->buckets[i];
0591
0592         if (ufid)
0593             hlist_for_each_entry_rcu(flow, head,
0594                          ufid_table.node[old_ver],
0595                          lockdep_ovsl_is_held())
0596                 ufid_table_instance_insert(new, flow);
0597         else
0598             hlist_for_each_entry_rcu(flow, head,
0599                          flow_table.node[old_ver],
0600                          lockdep_ovsl_is_held())
0601                 table_instance_insert(new, flow);
0602     }
0603 }
0604
0605 static struct table_instance *table_instance_rehash(struct table_instance *ti,
0606                             int n_buckets, bool ufid)
0607 {
0608     struct table_instance *new_ti;
0609
0610     new_ti = table_instance_alloc(n_buckets);
0611     if (!new_ti)
0612         return NULL;
0613
0614     flow_table_copy_flows(ti, new_ti, ufid);
0615
0616     return new_ti;
0617 }
0618
0619 int ovs_flow_tbl_flush(struct flow_table *flow_table)
0620 {
0621     struct table_instance *old_ti, *new_ti;
0622     struct table_instance *old_ufid_ti, *new_ufid_ti;
0623
0624     new_ti = table_instance_alloc(TBL_MIN_BUCKETS);
0625     if (!new_ti)
0626         return -ENOMEM;
0627     new_ufid_ti = table_instance_alloc(TBL_MIN_BUCKETS);
0628     if (!new_ufid_ti)
0629         goto err_free_ti;
0630
0631     old_ti = ovsl_dereference(flow_table->ti);
0632     old_ufid_ti = ovsl_dereference(flow_table->ufid_ti);
0633
0634     rcu_assign_pointer(flow_table->ti, new_ti);
0635     rcu_assign_pointer(flow_table->ufid_ti, new_ufid_ti);
0636     flow_table->last_rehash = jiffies;
0637
0638     table_instance_flow_flush(flow_table, old_ti, old_ufid_ti);
0639     table_instance_destroy(old_ti, old_ufid_ti);
0640     return 0;
0641
0642 err_free_ti:
0643     __table_instance_destroy(new_ti);
0644     return -ENOMEM;
0645 }
0646
0647 static u32 flow_hash(const struct sw_flow_key *key,
0648              const struct sw_flow_key_range *range)
0649 {
0650     const u32 *hash_key = (const u32 *)((const u8 *)key + range->start);
0651
0652     /* Make sure number of hash bytes are multiple of u32. */
0653     int hash_u32s = range_n_bytes(range) >> 2;
0654
0655     return jhash2(hash_key, hash_u32s, 0);
0656 }
0657
0658 static int flow_key_start(const struct sw_flow_key *key)
0659 {
0660     if (key->tun_proto)
0661         return 0;
0662     else
0663         return rounddown(offsetof(struct sw_flow_key, phy),
0664                  sizeof(long));
0665 }
0666
0667 static bool cmp_key(const struct sw_flow_key *key1,
0668             const struct sw_flow_key *key2,
0669             int key_start, int key_end)
0670 {
0671     const long *cp1 = (const long *)((const u8 *)key1 + key_start);
0672     const long *cp2 = (const long *)((const u8 *)key2 + key_start);
0673     int i;
0674
0675     for (i = key_start; i < key_end; i += sizeof(long))
0676         if (*cp1++ ^ *cp2++)
0677             return false;
0678
0679     return true;
0680 }
0681
0682 static bool flow_cmp_masked_key(const struct sw_flow *flow,
0683                 const struct sw_flow_key *key,
0684                 const struct sw_flow_key_range *range)
0685 {
0686     return cmp_key(&flow->key, key, range->start, range->end);
0687 }
0688
0689 static bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
0690                       const struct sw_flow_match *match)
0691 {
0692     struct sw_flow_key *key = match->key;
0693     int key_start = flow_key_start(key);
0694     int key_end = match->range.end;
0695
0696     BUG_ON(ovs_identifier_is_ufid(&flow->id));
0697     return cmp_key(flow->id.unmasked_key, key, key_start, key_end);
0698 }
0699
0700 static struct sw_flow *masked_flow_lookup(struct table_instance *ti,
0701                       const struct sw_flow_key *unmasked,
0702                       const struct sw_flow_mask *mask,
0703                       u32 *n_mask_hit)
0704 {
0705     struct sw_flow *flow;
0706     struct hlist_head *head;
0707     u32 hash;
0708     struct sw_flow_key masked_key;
0709
0710     ovs_flow_mask_key(&masked_key, unmasked, false, mask);
0711     hash = flow_hash(&masked_key, &mask->range);
0712     head = find_bucket(ti, hash);
0713     (*n_mask_hit)++;
0714
0715     hlist_for_each_entry_rcu(flow, head, flow_table.node[ti->node_ver],
0716                  lockdep_ovsl_is_held()) {
0717         if (flow->mask == mask && flow->flow_table.hash == hash &&
0718             flow_cmp_masked_key(flow, &masked_key, &mask->range))
0719             return flow;
0720     }
0721     return NULL;
0722 }
0723
0724 /* Flow lookup does full lookup on flow table. It starts with
0725  * mask from index passed in *index.
0726  * This function MUST be called with BH disabled due to the use
0727  * of CPU specific variables.
0728  */
0729 static struct sw_flow *flow_lookup(struct flow_table *tbl,
0730                    struct table_instance *ti,
0731                    struct mask_array *ma,
0732                    const struct sw_flow_key *key,
0733                    u32 *n_mask_hit,
0734                    u32 *n_cache_hit,
0735                    u32 *index)
0736 {
0737     struct mask_array_stats *stats = this_cpu_ptr(ma->masks_usage_stats);
0738     struct sw_flow *flow;
0739     struct sw_flow_mask *mask;
0740     int i;
0741
0742     if (likely(*index < ma->max)) {
0743         mask = rcu_dereference_ovsl(ma->masks[*index]);
0744         if (mask) {
0745             flow = masked_flow_lookup(ti, key, mask, n_mask_hit);
0746             if (flow) {
0747                 u64_stats_update_begin(&stats->syncp);
0748                 stats->usage_cntrs[*index]++;
0749                 u64_stats_update_end(&stats->syncp);
0750                 (*n_cache_hit)++;
0751                 return flow;
0752             }
0753         }
0754     }
0755
0756     for (i = 0; i < ma->max; i++)  {
0757
0758         if (i == *index)
0759             continue;
0760
0761         mask = rcu_dereference_ovsl(ma->masks[i]);
0762         if (unlikely(!mask))
0763             break;
0764
0765         flow = masked_flow_lookup(ti, key, mask, n_mask_hit);
0766         if (flow) { /* Found */
0767             *index = i;
0768             u64_stats_update_begin(&stats->syncp);
0769             stats->usage_cntrs[*index]++;
0770             u64_stats_update_end(&stats->syncp);
0771             return flow;
0772         }
0773     }
0774
0775     return NULL;
0776 }
0777
0778 /*
0779  * mask_cache maps flow to probable mask. This cache is not tightly
0780  * coupled cache, It means updates to  mask list can result in inconsistent
0781  * cache entry in mask cache.
0782  * This is per cpu cache and is divided in MC_HASH_SEGS segments.
0783  * In case of a hash collision the entry is hashed in next segment.
0784  * */
0785 struct sw_flow *ovs_flow_tbl_lookup_stats(struct flow_table *tbl,
0786                       const struct sw_flow_key *key,
0787                       u32 skb_hash,
0788                       u32 *n_mask_hit,
0789                       u32 *n_cache_hit)
0790 {
0791     struct mask_cache *mc = rcu_dereference(tbl->mask_cache);
0792     struct mask_array *ma = rcu_dereference(tbl->mask_array);
0793     struct table_instance *ti = rcu_dereference(tbl->ti);
0794     struct mask_cache_entry *entries, *ce;
0795     struct sw_flow *flow;
0796     u32 hash;
0797     int seg;
0798
0799     *n_mask_hit = 0;
0800     *n_cache_hit = 0;
0801     if (unlikely(!skb_hash || mc->cache_size == 0)) {
0802         u32 mask_index = 0;
0803         u32 cache = 0;
0804
0805         return flow_lookup(tbl, ti, ma, key, n_mask_hit, &cache,
0806                    &mask_index);
0807     }
0808
0809     /* Pre and post recirulation flows usually have the same skb_hash
0810      * value. To avoid hash collisions, rehash the 'skb_hash' with
0811      * 'recirc_id'.  */
0812     if (key->recirc_id)
0813         skb_hash = jhash_1word(skb_hash, key->recirc_id);
0814
0815     ce = NULL;
0816     hash = skb_hash;
0817     entries = this_cpu_ptr(mc->mask_cache);
0818
0819     /* Find the cache entry 'ce' to operate on. */
0820     for (seg = 0; seg < MC_HASH_SEGS; seg++) {
0821         int index = hash & (mc->cache_size - 1);
0822         struct mask_cache_entry *e;
0823
0824         e = &entries[index];
0825         if (e->skb_hash == skb_hash) {
0826             flow = flow_lookup(tbl, ti, ma, key, n_mask_hit,
0827                        n_cache_hit, &e->mask_index);
0828             if (!flow)
0829                 e->skb_hash = 0;
0830             return flow;
0831         }
0832
0833         if (!ce || e->skb_hash < ce->skb_hash)
0834             ce = e;  /* A better replacement cache candidate. */
0835
0836         hash >>= MC_HASH_SHIFT;
0837     }
0838
0839     /* Cache miss, do full lookup. */
0840     flow = flow_lookup(tbl, ti, ma, key, n_mask_hit, n_cache_hit,
0841                &ce->mask_index);
0842     if (flow)
0843         ce->skb_hash = skb_hash;
0844
0845     *n_cache_hit = 0;
0846     return flow;
0847 }
0848
0849 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
0850                     const struct sw_flow_key *key)
0851 {
0852     struct table_instance *ti = rcu_dereference_ovsl(tbl->ti);
0853     struct mask_array *ma = rcu_dereference_ovsl(tbl->mask_array);
0854     u32 __always_unused n_mask_hit;
0855     u32 __always_unused n_cache_hit;
0856     struct sw_flow *flow;
0857     u32 index = 0;
0858
0859     /* This function gets called trough the netlink interface and therefore
0860      * is preemptible. However, flow_lookup() function needs to be called
0861      * with BH disabled due to CPU specific variables.
0862      */
0863     local_bh_disable();
0864     flow = flow_lookup(tbl, ti, ma, key, &n_mask_hit, &n_cache_hit, &index);
0865     local_bh_enable();
0866     return flow;
0867 }
0868
0869 struct sw_flow *ovs_flow_tbl_lookup_exact(struct flow_table *tbl,
0870                       const struct sw_flow_match *match)
0871 {
0872     struct mask_array *ma = ovsl_dereference(tbl->mask_array);
0873     int i;
0874
0875     /* Always called under ovs-mutex. */
0876     for (i = 0; i < ma->max; i++) {
0877         struct table_instance *ti = rcu_dereference_ovsl(tbl->ti);
0878         u32 __always_unused n_mask_hit;
0879         struct sw_flow_mask *mask;
0880         struct sw_flow *flow;
0881
0882         mask = ovsl_dereference(ma->masks[i]);
0883         if (!mask)
0884             continue;
0885
0886         flow = masked_flow_lookup(ti, match->key, mask, &n_mask_hit);
0887         if (flow && ovs_identifier_is_key(&flow->id) &&
0888             ovs_flow_cmp_unmasked_key(flow, match)) {
0889             return flow;
0890         }
0891     }
0892
0893     return NULL;
0894 }
0895
0896 static u32 ufid_hash(const struct sw_flow_id *sfid)
0897 {
0898     return jhash(sfid->ufid, sfid->ufid_len, 0);
0899 }
0900
0901 static bool ovs_flow_cmp_ufid(const struct sw_flow *flow,
0902                   const struct sw_flow_id *sfid)
0903 {
0904     if (flow->id.ufid_len != sfid->ufid_len)
0905         return false;
0906
0907     return !memcmp(flow->id.ufid, sfid->ufid, sfid->ufid_len);
0908 }
0909
0910 bool ovs_flow_cmp(const struct sw_flow *flow,
0911           const struct sw_flow_match *match)
0912 {
0913     if (ovs_identifier_is_ufid(&flow->id))
0914         return flow_cmp_masked_key(flow, match->key, &match->range);
0915
0916     return ovs_flow_cmp_unmasked_key(flow, match);
0917 }
0918
0919 struct sw_flow *ovs_flow_tbl_lookup_ufid(struct flow_table *tbl,
0920                      const struct sw_flow_id *ufid)
0921 {
0922     struct table_instance *ti = rcu_dereference_ovsl(tbl->ufid_ti);
0923     struct sw_flow *flow;
0924     struct hlist_head *head;
0925     u32 hash;
0926
0927     hash = ufid_hash(ufid);
0928     head = find_bucket(ti, hash);
0929     hlist_for_each_entry_rcu(flow, head, ufid_table.node[ti->node_ver],
0930                  lockdep_ovsl_is_held()) {
0931         if (flow->ufid_table.hash == hash &&
0932             ovs_flow_cmp_ufid(flow, ufid))
0933             return flow;
0934     }
0935     return NULL;
0936 }
0937
0938 int ovs_flow_tbl_num_masks(const struct flow_table *table)
0939 {
0940     struct mask_array *ma = rcu_dereference_ovsl(table->mask_array);
0941     return READ_ONCE(ma->count);
0942 }
0943
0944 u32 ovs_flow_tbl_masks_cache_size(const struct flow_table *table)
0945 {
0946     struct mask_cache *mc = rcu_dereference_ovsl(table->mask_cache);
0947
0948     return READ_ONCE(mc->cache_size);
0949 }
0950
0951 static struct table_instance *table_instance_expand(struct table_instance *ti,
0952                             bool ufid)
0953 {
0954     return table_instance_rehash(ti, ti->n_buckets * 2, ufid);
0955 }
0956
0957 /* Must be called with OVS mutex held. */
0958 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
0959 {
0960     struct table_instance *ti = ovsl_dereference(table->ti);
0961     struct table_instance *ufid_ti = ovsl_dereference(table->ufid_ti);
0962
0963     BUG_ON(table->count == 0);
0964     table_instance_flow_free(table, ti, ufid_ti, flow);
0965 }
0966
0967 static struct sw_flow_mask *mask_alloc(void)
0968 {
0969     struct sw_flow_mask *mask;
0970
0971     mask = kmalloc(sizeof(*mask), GFP_KERNEL);
0972     if (mask)
0973         mask->ref_count = 1;
0974
0975     return mask;
0976 }
0977
0978 static bool mask_equal(const struct sw_flow_mask *a,
0979                const struct sw_flow_mask *b)
0980 {
0981     const u8 *a_ = (const u8 *)&a->key + a->range.start;
0982     const u8 *b_ = (const u8 *)&b->key + b->range.start;
0983
0984     return  (a->range.end == b->range.end)
0985         && (a->range.start == b->range.start)
0986         && (memcmp(a_, b_, range_n_bytes(&a->range)) == 0);
0987 }
0988
0989 static struct sw_flow_mask *flow_mask_find(const struct flow_table *tbl,
0990                        const struct sw_flow_mask *mask)
0991 {
0992     struct mask_array *ma;
0993     int i;
0994
0995     ma = ovsl_dereference(tbl->mask_array);
0996     for (i = 0; i < ma->max; i++) {
0997         struct sw_flow_mask *t;
0998         t = ovsl_dereference(ma->masks[i]);
0999
1000         if (t && mask_equal(mask, t))
1001             return t;
1002     }
1003
1004     return NULL;
1005 }
1006
1007 /* Add 'mask' into the mask list, if it is not already there. */
1008 static int flow_mask_insert(struct flow_table *tbl, struct sw_flow *flow,
1009                 const struct sw_flow_mask *new)
1010 {
1011     struct sw_flow_mask *mask;
1012
1013     mask = flow_mask_find(tbl, new);
1014     if (!mask) {
1015         /* Allocate a new mask if none exsits. */
1016         mask = mask_alloc();
1017         if (!mask)
1018             return -ENOMEM;
1019         mask->key = new->key;
1020         mask->range = new->range;
1021
1022         /* Add mask to mask-list. */
1023         if (tbl_mask_array_add_mask(tbl, mask)) {
1024             kfree(mask);
1025             return -ENOMEM;
1026         }
1027     } else {
1028         BUG_ON(!mask->ref_count);
1029         mask->ref_count++;
1030     }
1031
1032     flow->mask = mask;
1033     return 0;
1034 }
1035
1036 /* Must be called with OVS mutex held. */
1037 static void flow_key_insert(struct flow_table *table, struct sw_flow *flow)
1038 {
1039     struct table_instance *new_ti = NULL;
1040     struct table_instance *ti;
1041
1042     flow->flow_table.hash = flow_hash(&flow->key, &flow->mask->range);
1043     ti = ovsl_dereference(table->ti);
1044     table_instance_insert(ti, flow);
1045     table->count++;
1046
1047     /* Expand table, if necessary, to make room. */
1048     if (table->count > ti->n_buckets)
1049         new_ti = table_instance_expand(ti, false);
1050     else if (time_after(jiffies, table->last_rehash + REHASH_INTERVAL))
1051         new_ti = table_instance_rehash(ti, ti->n_buckets, false);
1052
1053     if (new_ti) {
1054         rcu_assign_pointer(table->ti, new_ti);
1055         call_rcu(&ti->rcu, flow_tbl_destroy_rcu_cb);
1056         table->last_rehash = jiffies;
1057     }
1058 }
1059
1060 /* Must be called with OVS mutex held. */
1061 static void flow_ufid_insert(struct flow_table *table, struct sw_flow *flow)
1062 {
1063     struct table_instance *ti;
1064
1065     flow->ufid_table.hash = ufid_hash(&flow->id);
1066     ti = ovsl_dereference(table->ufid_ti);
1067     ufid_table_instance_insert(ti, flow);
1068     table->ufid_count++;
1069
1070     /* Expand table, if necessary, to make room. */
1071     if (table->ufid_count > ti->n_buckets) {
1072         struct table_instance *new_ti;
1073
1074         new_ti = table_instance_expand(ti, true);
1075         if (new_ti) {
1076             rcu_assign_pointer(table->ufid_ti, new_ti);
1077             call_rcu(&ti->rcu, flow_tbl_destroy_rcu_cb);
1078         }
1079     }
1080 }
1081
1082 /* Must be called with OVS mutex held. */
1083 int ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
1084             const struct sw_flow_mask *mask)
1085 {
1086     int err;
1087
1088     err = flow_mask_insert(table, flow, mask);
1089     if (err)
1090         return err;
1091     flow_key_insert(table, flow);
1092     if (ovs_identifier_is_ufid(&flow->id))
1093         flow_ufid_insert(table, flow);
1094
1095     return 0;
1096 }
1097
1098 static int compare_mask_and_count(const void *a, const void *b)
1099 {
1100     const struct mask_count *mc_a = a;
1101     const struct mask_count *mc_b = b;
1102
1103     return (s64)mc_b->counter - (s64)mc_a->counter;
1104 }
1105
1106 /* Must be called with OVS mutex held. */
1107 void ovs_flow_masks_rebalance(struct flow_table *table)
1108 {
1109     struct mask_array *ma = rcu_dereference_ovsl(table->mask_array);
1110     struct mask_count *masks_and_count;
1111     struct mask_array *new;
1112     int masks_entries = 0;
1113     int i;
1114
1115     /* Build array of all current entries with use counters. */
1116     masks_and_count = kmalloc_array(ma->max, sizeof(*masks_and_count),
1117                     GFP_KERNEL);
1118     if (!masks_and_count)
1119         return;
1120
1121     for (i = 0; i < ma->max; i++) {
1122         struct sw_flow_mask *mask;
1123         int cpu;
1124
1125         mask = rcu_dereference_ovsl(ma->masks[i]);
1126         if (unlikely(!mask))
1127             break;
1128
1129         masks_and_count[i].index = i;
1130         masks_and_count[i].counter = 0;
1131
1132         for_each_possible_cpu(cpu) {
1133             struct mask_array_stats *stats;
1134             unsigned int start;
1135             u64 counter;
1136
1137             stats = per_cpu_ptr(ma->masks_usage_stats, cpu);
1138             do {
1139                 start = u64_stats_fetch_begin_irq(&stats->syncp);
1140                 counter = stats->usage_cntrs[i];
1141             } while (u64_stats_fetch_retry_irq(&stats->syncp,
1142                                start));
1143
1144             masks_and_count[i].counter += counter;
1145         }
1146
1147         /* Subtract the zero count value. */
1148         masks_and_count[i].counter -= ma->masks_usage_zero_cntr[i];
1149
1150         /* Rather than calling tbl_mask_array_reset_counters()
1151          * below when no change is needed, do it inline here.
1152          */
1153         ma->masks_usage_zero_cntr[i] += masks_and_count[i].counter;
1154     }
1155
1156     if (i == 0)
1157         goto free_mask_entries;
1158
1159     /* Sort the entries */
1160     masks_entries = i;
1161     sort(masks_and_count, masks_entries, sizeof(*masks_and_count),
1162          compare_mask_and_count, NULL);
1163
1164     /* If the order is the same, nothing to do... */
1165     for (i = 0; i < masks_entries; i++) {
1166         if (i != masks_and_count[i].index)
1167             break;
1168     }
1169     if (i == masks_entries)
1170         goto free_mask_entries;
1171
1172     /* Rebuilt the new list in order of usage. */
1173     new = tbl_mask_array_alloc(ma->max);
1174     if (!new)
1175         goto free_mask_entries;
1176
1177     for (i = 0; i < masks_entries; i++) {
1178         int index = masks_and_count[i].index;
1179
1180         if (ovsl_dereference(ma->masks[index]))
1181             new->masks[new->count++] = ma->masks[index];
1182     }
1183
1184     rcu_assign_pointer(table->mask_array, new);
1185     call_rcu(&ma->rcu, mask_array_rcu_cb);
1186
1187 free_mask_entries:
1188     kfree(masks_and_count);
1189 }
1190
1191 /* Initializes the flow module.
1192  * Returns zero if successful or a negative error code. */
1193 int ovs_flow_init(void)
1194 {
1195     BUILD_BUG_ON(__alignof__(struct sw_flow_key) % __alignof__(long));
1196     BUILD_BUG_ON(sizeof(struct sw_flow_key) % sizeof(long));
1197
1198     flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow)
1199                        + (nr_cpu_ids
1200                       * sizeof(struct sw_flow_stats *)),
1201                        0, 0, NULL);
1202     if (flow_cache == NULL)
1203         return -ENOMEM;
1204
1205     flow_stats_cache
1206         = kmem_cache_create("sw_flow_stats", sizeof(struct sw_flow_stats),
1207                     0, SLAB_HWCACHE_ALIGN, NULL);
1208     if (flow_stats_cache == NULL) {
1209         kmem_cache_destroy(flow_cache);
1210         flow_cache = NULL;
1211         return -ENOMEM;
1212     }
1213
1214     return 0;
1215 }
1216
1217 /* Uninitializes the flow module. */
1218 void ovs_flow_exit(void)
1219 {
1220     kmem_cache_destroy(flow_stats_cache);
1221     kmem_cache_destroy(flow_cache);
1222 }