Commit | Line | Data |
---|---|---|
5ab11c98 PE |
1 | #ifndef __NET_FRAG_H__ |
2 | #define __NET_FRAG_H__ | |
3 | ||
6d7b857d JDB |
4 | #include <linux/percpu_counter.h> |
5 | ||
ac18e750 | 6 | struct netns_frags { |
6d7b857d JDB |
7 | /* The percpu_counter "mem" need to be cacheline aligned. |
8 | * mem.count must not share cacheline with other writers | |
cd39a789 | 9 | */ |
6d7b857d JDB |
10 | struct percpu_counter mem ____cacheline_aligned_in_smp; |
11 | ||
b2fd5321 PE |
12 | /* sysctls */ |
13 | int timeout; | |
e31e0bdc PE |
14 | int high_thresh; |
15 | int low_thresh; | |
ac18e750 PE |
16 | }; |
17 | ||
5ab11c98 | 18 | struct inet_frag_queue { |
5ab11c98 | 19 | spinlock_t lock; |
5ab11c98 | 20 | struct timer_list timer; /* when will this queue expire? */ |
6e34a8b3 JDB |
21 | struct hlist_node list; |
22 | atomic_t refcnt; | |
5ab11c98 | 23 | struct sk_buff *fragments; /* list of received fragments */ |
d6bebca9 | 24 | struct sk_buff *fragments_tail; |
5ab11c98 PE |
25 | ktime_t stamp; |
26 | int len; /* total length of orig datagram */ | |
27 | int meat; | |
28 | __u8 last_in; /* first/last segment arrived? */ | |
29 | ||
b13d3cbf | 30 | #define INET_FRAG_EVICTED 8 |
bc578a54 JP |
31 | #define INET_FRAG_COMPLETE 4 |
32 | #define INET_FRAG_FIRST_IN 2 | |
33 | #define INET_FRAG_LAST_IN 1 | |
5f2d04f1 PM |
34 | |
35 | u16 max_size; | |
6e34a8b3 JDB |
36 | |
37 | struct netns_frags *net; | |
5ab11c98 PE |
38 | }; |
39 | ||
a4c4009f | 40 | #define INETFRAGS_HASHSZ 1024 |
7eb95156 | 41 | |
5a3da1fe HFS |
42 | /* averaged: |
43 | * max_depth = default ipfrag_high_thresh / INETFRAGS_HASHSZ / | |
44 | * rounded up (SKB_TRUELEN(0) + sizeof(struct ipq or | |
45 | * struct frag_queue)) | |
46 | */ | |
b13d3cbf | 47 | #define INETFRAGS_MAXDEPTH 128 |
5a3da1fe | 48 | |
19952cc4 JDB |
49 | struct inet_frag_bucket { |
50 | struct hlist_head chain; | |
51 | spinlock_t chain_lock; | |
52 | }; | |
53 | ||
7eb95156 | 54 | struct inet_frags { |
19952cc4 | 55 | struct inet_frag_bucket hash[INETFRAGS_HASHSZ]; |
5f8e1e8b JDB |
56 | /* This rwlock is a global lock (seperate per IPv4, IPv6 and |
57 | * netfilter). Important to keep this on a seperate cacheline. | |
19952cc4 | 58 | * Its primarily a rebuild protection rwlock. |
5f8e1e8b JDB |
59 | */ |
60 | rwlock_t lock ____cacheline_aligned_in_smp; | |
3b4bc4a2 | 61 | int secret_interval; |
7eb95156 | 62 | struct timer_list secret_timer; |
7088ad74 | 63 | |
b13d3cbf FW |
64 | struct work_struct frags_work; |
65 | unsigned int next_bucket; | |
66 | ||
7088ad74 HFS |
67 | /* The first call to hashfn is responsible to initialize |
68 | * rnd. This is best done with net_get_random_once. | |
69 | */ | |
5f8e1e8b JDB |
70 | u32 rnd; |
71 | int qsize; | |
321a3a99 | 72 | |
36c77782 FW |
73 | unsigned int (*hashfn)(const struct inet_frag_queue *); |
74 | bool (*match)(const struct inet_frag_queue *q, | |
75 | const void *arg); | |
c6fda282 | 76 | void (*constructor)(struct inet_frag_queue *q, |
36c77782 | 77 | const void *arg); |
1e4b8287 PE |
78 | void (*destructor)(struct inet_frag_queue *); |
79 | void (*skb_free)(struct sk_buff *); | |
e521db9d | 80 | void (*frag_expire)(unsigned long data); |
7eb95156 PE |
81 | }; |
82 | ||
83 | void inet_frags_init(struct inet_frags *); | |
84 | void inet_frags_fini(struct inet_frags *); | |
85 | ||
e5a2bb84 | 86 | void inet_frags_init_net(struct netns_frags *nf); |
81566e83 | 87 | void inet_frags_exit_net(struct netns_frags *nf, struct inet_frags *f); |
e5a2bb84 | 88 | |
277e650d | 89 | void inet_frag_kill(struct inet_frag_queue *q, struct inet_frags *f); |
3fd588eb | 90 | void inet_frag_destroy(struct inet_frag_queue *q, struct inet_frags *f); |
ac18e750 | 91 | struct inet_frag_queue *inet_frag_find(struct netns_frags *nf, |
56bca31f HE |
92 | struct inet_frags *f, void *key, unsigned int hash) |
93 | __releases(&f->lock); | |
5a3da1fe HFS |
94 | void inet_frag_maybe_warn_overflow(struct inet_frag_queue *q, |
95 | const char *prefix); | |
277e650d | 96 | |
762cc408 PE |
97 | static inline void inet_frag_put(struct inet_frag_queue *q, struct inet_frags *f) |
98 | { | |
99 | if (atomic_dec_and_test(&q->refcnt)) | |
3fd588eb | 100 | inet_frag_destroy(q, f); |
762cc408 PE |
101 | } |
102 | ||
d433673e JDB |
103 | /* Memory Tracking Functions. */ |
104 | ||
6d7b857d JDB |
105 | /* The default percpu_counter batch size is not big enough to scale to |
106 | * fragmentation mem acct sizes. | |
107 | * The mem size of a 64K fragment is approx: | |
108 | * (44 fragments * 2944 truesize) + frag_queue struct(200) = 129736 bytes | |
109 | */ | |
110 | static unsigned int frag_percpu_counter_batch = 130000; | |
111 | ||
d433673e JDB |
112 | static inline int frag_mem_limit(struct netns_frags *nf) |
113 | { | |
6d7b857d | 114 | return percpu_counter_read(&nf->mem); |
d433673e JDB |
115 | } |
116 | ||
117 | static inline void sub_frag_mem_limit(struct inet_frag_queue *q, int i) | |
118 | { | |
6d7b857d | 119 | __percpu_counter_add(&q->net->mem, -i, frag_percpu_counter_batch); |
d433673e JDB |
120 | } |
121 | ||
122 | static inline void add_frag_mem_limit(struct inet_frag_queue *q, int i) | |
123 | { | |
6d7b857d | 124 | __percpu_counter_add(&q->net->mem, i, frag_percpu_counter_batch); |
d433673e JDB |
125 | } |
126 | ||
127 | static inline void init_frag_mem_limit(struct netns_frags *nf) | |
128 | { | |
6d7b857d | 129 | percpu_counter_init(&nf->mem, 0); |
d433673e JDB |
130 | } |
131 | ||
36c77782 | 132 | static inline unsigned int sum_frag_mem_limit(struct netns_frags *nf) |
d433673e | 133 | { |
36c77782 | 134 | unsigned int res; |
4cfb0485 ED |
135 | |
136 | local_bh_disable(); | |
137 | res = percpu_counter_sum_positive(&nf->mem); | |
138 | local_bh_enable(); | |
139 | ||
140 | return res; | |
d433673e JDB |
141 | } |
142 | ||
be991971 HFS |
143 | /* RFC 3168 support : |
144 | * We want to check ECN values of all fragments, do detect invalid combinations. | |
145 | * In ipq->ecn, we store the OR value of each ip4_frag_ecn() fragment value. | |
146 | */ | |
147 | #define IPFRAG_ECN_NOT_ECT 0x01 /* one frag had ECN_NOT_ECT */ | |
148 | #define IPFRAG_ECN_ECT_1 0x02 /* one frag had ECN_ECT_1 */ | |
149 | #define IPFRAG_ECN_ECT_0 0x04 /* one frag had ECN_ECT_0 */ | |
150 | #define IPFRAG_ECN_CE 0x08 /* one frag had ECN_CE */ | |
151 | ||
152 | extern const u8 ip_frag_ecn_table[16]; | |
153 | ||
5ab11c98 | 154 | #endif |