1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef __NET_FRAG_H__
3 #define __NET_FRAG_H__
4
5 #include <linux/rhashtable-types.h>
6 #include <linux/completion.h>
7
8 /* Per netns frag queues directory */
9 struct fqdir {
10 /* sysctls */
11 long high_thresh;
12 long low_thresh;
13 int timeout;
14 int max_dist;
15 struct inet_frags *f;
16 struct net *net;
17 bool dead;
18
19 struct rhashtable rhashtable ____cacheline_aligned_in_smp;
20
21 /* Keep atomic mem on separate cachelines in structs that include it */
22 atomic_long_t mem ____cacheline_aligned_in_smp;
23 struct work_struct destroy_work;
24 struct llist_node free_list;
25 };
26
27 /**
28 * fragment queue flags
29 *
30 * @INET_FRAG_FIRST_IN: first fragment has arrived
31 * @INET_FRAG_LAST_IN: final fragment has arrived
32 * @INET_FRAG_COMPLETE: frag queue has been processed and is due for destruction
33 * @INET_FRAG_HASH_DEAD: inet_frag_kill() has not removed fq from rhashtable
34 */
35 enum {
36 INET_FRAG_FIRST_IN = BIT(0),
37 INET_FRAG_LAST_IN = BIT(1),
38 INET_FRAG_COMPLETE = BIT(2),
39 INET_FRAG_HASH_DEAD = BIT(3),
40 };
41
42 struct frag_v4_compare_key {
43 __be32 saddr;
44 __be32 daddr;
45 u32 user;
46 u32 vif;
47 __be16 id;
48 u16 protocol;
49 };
50
51 struct frag_v6_compare_key {
52 struct in6_addr saddr;
53 struct in6_addr daddr;
54 u32 user;
55 __be32 id;
56 u32 iif;
57 };
58
59 /**
60 * struct inet_frag_queue - fragment queue
61 *
62 * @node: rhash node
63 * @key: keys identifying this frag.
64 * @timer: queue expiration timer
65 * @lock: spinlock protecting this frag
66 * @refcnt: reference count of the queue
67 * @rb_fragments: received fragments rb-tree root
68 * @fragments_tail: received fragments tail
69 * @last_run_head: the head of the last "run". see ip_fragment.c
70 * @stamp: timestamp of the last received fragment
71 * @len: total length of the original datagram
72 * @meat: length of received fragments so far
73 * @mono_delivery_time: stamp has a mono delivery time (EDT)
74 * @flags: fragment queue flags
75 * @max_size: maximum received fragment size
76 * @fqdir: pointer to struct fqdir
77 * @rcu: rcu head for freeing deferall
78 */
79 struct inet_frag_queue {
80 struct rhash_head node;
81 union {
82 struct frag_v4_compare_key v4;
83 struct frag_v6_compare_key v6;
84 } key;
85 struct timer_list timer;
86 spinlock_t lock;
87 refcount_t refcnt;
88 struct rb_root rb_fragments;
89 struct sk_buff *fragments_tail;
90 struct sk_buff *last_run_head;
91 ktime_t stamp;
92 int len;
93 int meat;
94 u8 mono_delivery_time;
95 __u8 flags;
96 u16 max_size;
97 struct fqdir *fqdir;
98 struct rcu_head rcu;
99 };
100
101 struct inet_frags {
102 unsigned int qsize;
103
104 void (*constructor)(struct inet_frag_queue *q,
105 const void *arg);
106 void (*destructor)(struct inet_frag_queue *);
107 void (*frag_expire)(struct timer_list *t);
108 struct kmem_cache *frags_cachep;
109 const char *frags_cache_name;
110 struct rhashtable_params rhash_params;
111 refcount_t refcnt;
112 struct completion completion;
113 };
114
115 int inet_frags_init(struct inet_frags *);
116 void inet_frags_fini(struct inet_frags *);
117
118 int fqdir_init(struct fqdir **fqdirp, struct inet_frags *f, struct net *net);
119
fqdir_pre_exit(struct fqdir * fqdir)120 static inline void fqdir_pre_exit(struct fqdir *fqdir)
121 {
122 /* Prevent creation of new frags.
123 * Pairs with READ_ONCE() in inet_frag_find().
124 */
125 WRITE_ONCE(fqdir->high_thresh, 0);
126
127 /* Pairs with READ_ONCE() in inet_frag_kill(), ip_expire()
128 * and ip6frag_expire_frag_queue().
129 */
130 WRITE_ONCE(fqdir->dead, true);
131 }
132 void fqdir_exit(struct fqdir *fqdir);
133
134 void inet_frag_kill(struct inet_frag_queue *q);
135 void inet_frag_destroy(struct inet_frag_queue *q);
136 struct inet_frag_queue *inet_frag_find(struct fqdir *fqdir, void *key);
137
138 /* Free all skbs in the queue; return the sum of their truesizes. */
139 unsigned int inet_frag_rbtree_purge(struct rb_root *root);
140
inet_frag_put(struct inet_frag_queue * q)141 static inline void inet_frag_put(struct inet_frag_queue *q)
142 {
143 if (refcount_dec_and_test(&q->refcnt))
144 inet_frag_destroy(q);
145 }
146
147 /* Memory Tracking Functions. */
148
frag_mem_limit(const struct fqdir * fqdir)149 static inline long frag_mem_limit(const struct fqdir *fqdir)
150 {
151 return atomic_long_read(&fqdir->mem);
152 }
153
sub_frag_mem_limit(struct fqdir * fqdir,long val)154 static inline void sub_frag_mem_limit(struct fqdir *fqdir, long val)
155 {
156 atomic_long_sub(val, &fqdir->mem);
157 }
158
add_frag_mem_limit(struct fqdir * fqdir,long val)159 static inline void add_frag_mem_limit(struct fqdir *fqdir, long val)
160 {
161 atomic_long_add(val, &fqdir->mem);
162 }
163
164 /* RFC 3168 support :
165 * We want to check ECN values of all fragments, do detect invalid combinations.
166 * In ipq->ecn, we store the OR value of each ip4_frag_ecn() fragment value.
167 */
168 #define IPFRAG_ECN_NOT_ECT 0x01 /* one frag had ECN_NOT_ECT */
169 #define IPFRAG_ECN_ECT_1 0x02 /* one frag had ECN_ECT_1 */
170 #define IPFRAG_ECN_ECT_0 0x04 /* one frag had ECN_ECT_0 */
171 #define IPFRAG_ECN_CE 0x08 /* one frag had ECN_CE */
172
173 extern const u8 ip_frag_ecn_table[16];
174
175 /* Return values of inet_frag_queue_insert() */
176 #define IPFRAG_OK 0
177 #define IPFRAG_DUP 1
178 #define IPFRAG_OVERLAP 2
179 int inet_frag_queue_insert(struct inet_frag_queue *q, struct sk_buff *skb,
180 int offset, int end);
181 void *inet_frag_reasm_prepare(struct inet_frag_queue *q, struct sk_buff *skb,
182 struct sk_buff *parent);
183 void inet_frag_reasm_finish(struct inet_frag_queue *q, struct sk_buff *head,
184 void *reasm_data, bool try_coalesce);
185 struct sk_buff *inet_frag_pull_head(struct inet_frag_queue *q);
186
187 #endif
188