Kyle Swenson | 8d8f654 | 2021-03-15 11:02:55 -0600 | [diff] [blame^] | 1 | #ifndef __NET_FRAG_H__ |
| 2 | #define __NET_FRAG_H__ |
| 3 | |
| 4 | struct netns_frags { |
| 5 | /* Keep atomic mem on separate cachelines in structs that include it */ |
| 6 | atomic_t mem ____cacheline_aligned_in_smp; |
| 7 | /* sysctls */ |
| 8 | int timeout; |
| 9 | int high_thresh; |
| 10 | int low_thresh; |
| 11 | }; |
| 12 | |
| 13 | /** |
| 14 | * fragment queue flags |
| 15 | * |
| 16 | * @INET_FRAG_FIRST_IN: first fragment has arrived |
| 17 | * @INET_FRAG_LAST_IN: final fragment has arrived |
| 18 | * @INET_FRAG_COMPLETE: frag queue has been processed and is due for destruction |
| 19 | */ |
| 20 | enum { |
| 21 | INET_FRAG_FIRST_IN = BIT(0), |
| 22 | INET_FRAG_LAST_IN = BIT(1), |
| 23 | INET_FRAG_COMPLETE = BIT(2), |
| 24 | }; |
| 25 | |
| 26 | /** |
| 27 | * struct inet_frag_queue - fragment queue |
| 28 | * |
| 29 | * @lock: spinlock protecting the queue |
| 30 | * @timer: queue expiration timer |
| 31 | * @list: hash bucket list |
| 32 | * @refcnt: reference count of the queue |
| 33 | * @fragments: received fragments head |
| 34 | * @fragments_tail: received fragments tail |
| 35 | * @stamp: timestamp of the last received fragment |
| 36 | * @len: total length of the original datagram |
| 37 | * @meat: length of received fragments so far |
| 38 | * @flags: fragment queue flags |
| 39 | * @max_size: maximum received fragment size |
| 40 | * @net: namespace that this frag belongs to |
| 41 | * @list_evictor: list of queues to forcefully evict (e.g. due to low memory) |
| 42 | */ |
| 43 | struct inet_frag_queue { |
| 44 | spinlock_t lock; |
| 45 | struct timer_list timer; |
| 46 | struct hlist_node list; |
| 47 | atomic_t refcnt; |
| 48 | struct sk_buff *fragments; |
| 49 | struct sk_buff *fragments_tail; |
| 50 | ktime_t stamp; |
| 51 | int len; |
| 52 | int meat; |
| 53 | __u8 flags; |
| 54 | u16 max_size; |
| 55 | struct netns_frags *net; |
| 56 | struct hlist_node list_evictor; |
| 57 | }; |
| 58 | |
| 59 | #define INETFRAGS_HASHSZ 1024 |
| 60 | |
| 61 | /* averaged: |
| 62 | * max_depth = default ipfrag_high_thresh / INETFRAGS_HASHSZ / |
| 63 | * rounded up (SKB_TRUELEN(0) + sizeof(struct ipq or |
| 64 | * struct frag_queue)) |
| 65 | */ |
| 66 | #define INETFRAGS_MAXDEPTH 128 |
| 67 | |
| 68 | struct inet_frag_bucket { |
| 69 | struct hlist_head chain; |
| 70 | spinlock_t chain_lock; |
| 71 | }; |
| 72 | |
| 73 | struct inet_frags { |
| 74 | struct inet_frag_bucket hash[INETFRAGS_HASHSZ]; |
| 75 | |
| 76 | struct work_struct frags_work; |
| 77 | unsigned int next_bucket; |
| 78 | unsigned long last_rebuild_jiffies; |
| 79 | bool rebuild; |
| 80 | |
| 81 | /* The first call to hashfn is responsible to initialize |
| 82 | * rnd. This is best done with net_get_random_once. |
| 83 | * |
| 84 | * rnd_seqlock is used to let hash insertion detect |
| 85 | * when it needs to re-lookup the hash chain to use. |
| 86 | */ |
| 87 | u32 rnd; |
| 88 | seqlock_t rnd_seqlock; |
| 89 | int qsize; |
| 90 | |
| 91 | unsigned int (*hashfn)(const struct inet_frag_queue *); |
| 92 | bool (*match)(const struct inet_frag_queue *q, |
| 93 | const void *arg); |
| 94 | void (*constructor)(struct inet_frag_queue *q, |
| 95 | const void *arg); |
| 96 | void (*destructor)(struct inet_frag_queue *); |
| 97 | void (*skb_free)(struct sk_buff *); |
| 98 | void (*frag_expire)(unsigned long data); |
| 99 | struct kmem_cache *frags_cachep; |
| 100 | const char *frags_cache_name; |
| 101 | }; |
| 102 | |
| 103 | int inet_frags_init(struct inet_frags *); |
| 104 | void inet_frags_fini(struct inet_frags *); |
| 105 | |
| 106 | static inline void inet_frags_init_net(struct netns_frags *nf) |
| 107 | { |
| 108 | atomic_set(&nf->mem, 0); |
| 109 | } |
| 110 | void inet_frags_exit_net(struct netns_frags *nf, struct inet_frags *f); |
| 111 | |
| 112 | void inet_frag_kill(struct inet_frag_queue *q, struct inet_frags *f); |
| 113 | void inet_frag_destroy(struct inet_frag_queue *q, struct inet_frags *f); |
| 114 | struct inet_frag_queue *inet_frag_find(struct netns_frags *nf, |
| 115 | struct inet_frags *f, void *key, unsigned int hash); |
| 116 | |
| 117 | void inet_frag_maybe_warn_overflow(struct inet_frag_queue *q, |
| 118 | const char *prefix); |
| 119 | |
| 120 | static inline void inet_frag_put(struct inet_frag_queue *q, struct inet_frags *f) |
| 121 | { |
| 122 | if (atomic_dec_and_test(&q->refcnt)) |
| 123 | inet_frag_destroy(q, f); |
| 124 | } |
| 125 | |
| 126 | static inline bool inet_frag_evicting(struct inet_frag_queue *q) |
| 127 | { |
| 128 | return !hlist_unhashed(&q->list_evictor); |
| 129 | } |
| 130 | |
| 131 | /* Memory Tracking Functions. */ |
| 132 | |
| 133 | static inline int frag_mem_limit(struct netns_frags *nf) |
| 134 | { |
| 135 | return atomic_read(&nf->mem); |
| 136 | } |
| 137 | |
| 138 | static inline void sub_frag_mem_limit(struct netns_frags *nf, int i) |
| 139 | { |
| 140 | atomic_sub(i, &nf->mem); |
| 141 | } |
| 142 | |
| 143 | static inline void add_frag_mem_limit(struct netns_frags *nf, int i) |
| 144 | { |
| 145 | atomic_add(i, &nf->mem); |
| 146 | } |
| 147 | |
| 148 | static inline int sum_frag_mem_limit(struct netns_frags *nf) |
| 149 | { |
| 150 | return atomic_read(&nf->mem); |
| 151 | } |
| 152 | |
| 153 | /* RFC 3168 support : |
| 154 | * We want to check ECN values of all fragments, do detect invalid combinations. |
| 155 | * In ipq->ecn, we store the OR value of each ip4_frag_ecn() fragment value. |
| 156 | */ |
| 157 | #define IPFRAG_ECN_NOT_ECT 0x01 /* one frag had ECN_NOT_ECT */ |
| 158 | #define IPFRAG_ECN_ECT_1 0x02 /* one frag had ECN_ECT_1 */ |
| 159 | #define IPFRAG_ECN_ECT_0 0x04 /* one frag had ECN_ECT_0 */ |
| 160 | #define IPFRAG_ECN_CE 0x08 /* one frag had ECN_CE */ |
| 161 | |
| 162 | extern const u8 ip_frag_ecn_table[16]; |
| 163 | |
| 164 | #endif |