4 #include <linux/percpu_counter.h>
7 /* The percpu_counter "mem" need to be cacheline aligned.
8 * mem.count must not share cacheline with other writers
10 struct percpu_counter mem ____cacheline_aligned_in_smp;
19 * fragment queue flags
21 * @INET_FRAG_FIRST_IN: first fragment has arrived
22 * @INET_FRAG_LAST_IN: final fragment has arrived
23 * @INET_FRAG_COMPLETE: frag queue has been processed and is due for destruction
24 * @INET_FRAG_EVICTED: frag queue is being evicted
27 INET_FRAG_FIRST_IN = BIT(0),
28 INET_FRAG_LAST_IN = BIT(1),
29 INET_FRAG_COMPLETE = BIT(2),
30 INET_FRAG_EVICTED = BIT(3)
34 * struct inet_frag_queue - fragment queue
36 * @lock: spinlock protecting the queue
37 * @timer: queue expiration timer
38 * @list: hash bucket list
39 * @refcnt: reference count of the queue
40 * @fragments: received fragments head
41 * @fragments_tail: received fragments tail
42 * @stamp: timestamp of the last received fragment
43 * @len: total length of the original datagram
44 * @meat: length of received fragments so far
45 * @flags: fragment queue flags
46 * @max_size: (ipv4 only) maximum received fragment size with IP_DF set
47 * @net: namespace that this frag belongs to
49 struct inet_frag_queue {
51 struct timer_list timer;
52 struct hlist_node list;
54 struct sk_buff *fragments;
55 struct sk_buff *fragments_tail;
61 struct netns_frags *net;
64 #define INETFRAGS_HASHSZ 1024
67 * max_depth = default ipfrag_high_thresh / INETFRAGS_HASHSZ /
68 * rounded up (SKB_TRUELEN(0) + sizeof(struct ipq or
71 #define INETFRAGS_MAXDEPTH 128
73 struct inet_frag_bucket {
74 struct hlist_head chain;
75 spinlock_t chain_lock;
79 struct inet_frag_bucket hash[INETFRAGS_HASHSZ];
81 struct work_struct frags_work;
82 unsigned int next_bucket;
83 unsigned long last_rebuild_jiffies;
86 /* The first call to hashfn is responsible to initialize
87 * rnd. This is best done with net_get_random_once.
89 * rnd_seqlock is used to let hash insertion detect
90 * when it needs to re-lookup the hash chain to use.
93 seqlock_t rnd_seqlock;
96 unsigned int (*hashfn)(const struct inet_frag_queue *);
97 bool (*match)(const struct inet_frag_queue *q,
99 void (*constructor)(struct inet_frag_queue *q,
101 void (*destructor)(struct inet_frag_queue *);
102 void (*skb_free)(struct sk_buff *);
103 void (*frag_expire)(unsigned long data);
104 struct kmem_cache *frags_cachep;
105 const char *frags_cache_name;
108 int inet_frags_init(struct inet_frags *);
109 void inet_frags_fini(struct inet_frags *);
111 void inet_frags_init_net(struct netns_frags *nf);
112 void inet_frags_exit_net(struct netns_frags *nf, struct inet_frags *f);
114 void inet_frag_kill(struct inet_frag_queue *q, struct inet_frags *f);
115 void inet_frag_destroy(struct inet_frag_queue *q, struct inet_frags *f);
116 struct inet_frag_queue *inet_frag_find(struct netns_frags *nf,
117 struct inet_frags *f, void *key, unsigned int hash);
119 void inet_frag_maybe_warn_overflow(struct inet_frag_queue *q,
122 static inline void inet_frag_put(struct inet_frag_queue *q, struct inet_frags *f)
124 if (atomic_dec_and_test(&q->refcnt))
125 inet_frag_destroy(q, f);
128 /* Memory Tracking Functions. */
130 /* The default percpu_counter batch size is not big enough to scale to
131 * fragmentation mem acct sizes.
132 * The mem size of a 64K fragment is approx:
133 * (44 fragments * 2944 truesize) + frag_queue struct(200) = 129736 bytes
135 static unsigned int frag_percpu_counter_batch = 130000;
137 static inline int frag_mem_limit(struct netns_frags *nf)
139 return percpu_counter_read(&nf->mem);
142 static inline void sub_frag_mem_limit(struct inet_frag_queue *q, int i)
144 __percpu_counter_add(&q->net->mem, -i, frag_percpu_counter_batch);
147 static inline void add_frag_mem_limit(struct inet_frag_queue *q, int i)
149 __percpu_counter_add(&q->net->mem, i, frag_percpu_counter_batch);
152 static inline void init_frag_mem_limit(struct netns_frags *nf)
154 percpu_counter_init(&nf->mem, 0, GFP_KERNEL);
157 static inline unsigned int sum_frag_mem_limit(struct netns_frags *nf)
162 res = percpu_counter_sum_positive(&nf->mem);
168 /* RFC 3168 support :
169 * We want to check ECN values of all fragments, do detect invalid combinations.
170 * In ipq->ecn, we store the OR value of each ip4_frag_ecn() fragment value.
172 #define IPFRAG_ECN_NOT_ECT 0x01 /* one frag had ECN_NOT_ECT */
173 #define IPFRAG_ECN_ECT_1 0x02 /* one frag had ECN_ECT_1 */
174 #define IPFRAG_ECN_ECT_0 0x04 /* one frag had ECN_ECT_0 */
175 #define IPFRAG_ECN_CE 0x08 /* one frag had ECN_CE */
177 extern const u8 ip_frag_ecn_table[16];