Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
tjh.dev
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef __FS_CEPH_MESSENGER_H
3#define __FS_CEPH_MESSENGER_H
4
5#include <linux/bvec.h>
6#include <linux/kref.h>
7#include <linux/mutex.h>
8#include <linux/net.h>
9#include <linux/radix-tree.h>
10#include <linux/uio.h>
11#include <linux/workqueue.h>
12#include <net/net_namespace.h>
13
14#include <linux/ceph/types.h>
15#include <linux/ceph/buffer.h>
16
17struct ceph_msg;
18struct ceph_connection;
19
20/*
21 * Ceph defines these callbacks for handling connection events.
22 */
23struct ceph_connection_operations {
24 struct ceph_connection *(*get)(struct ceph_connection *);
25 void (*put)(struct ceph_connection *);
26
27 /* handle an incoming message. */
28 void (*dispatch) (struct ceph_connection *con, struct ceph_msg *m);
29
30 /* authorize an outgoing connection */
31 struct ceph_auth_handshake *(*get_authorizer) (
32 struct ceph_connection *con,
33 int *proto, int force_new);
34 int (*add_authorizer_challenge)(struct ceph_connection *con,
35 void *challenge_buf,
36 int challenge_buf_len);
37 int (*verify_authorizer_reply) (struct ceph_connection *con);
38 int (*invalidate_authorizer)(struct ceph_connection *con);
39
40 /* there was some error on the socket (disconnect, whatever) */
41 void (*fault) (struct ceph_connection *con);
42
43 /* a remote host as terminated a message exchange session, and messages
44 * we sent (or they tried to send us) may be lost. */
45 void (*peer_reset) (struct ceph_connection *con);
46
47 struct ceph_msg * (*alloc_msg) (struct ceph_connection *con,
48 struct ceph_msg_header *hdr,
49 int *skip);
50
51 void (*reencode_message) (struct ceph_msg *msg);
52
53 int (*sign_message) (struct ceph_msg *msg);
54 int (*check_message_signature) (struct ceph_msg *msg);
55};
56
57/* use format string %s%d */
58#define ENTITY_NAME(n) ceph_entity_type_name((n).type), le64_to_cpu((n).num)
59
60struct ceph_messenger {
61 struct ceph_entity_inst inst; /* my name+address */
62 struct ceph_entity_addr my_enc_addr;
63
64 atomic_t stopping;
65 possible_net_t net;
66
67 /*
68 * the global_seq counts connections i (attempt to) initiate
69 * in order to disambiguate certain connect race conditions.
70 */
71 u32 global_seq;
72 spinlock_t global_seq_lock;
73};
74
75enum ceph_msg_data_type {
76 CEPH_MSG_DATA_NONE, /* message contains no data payload */
77 CEPH_MSG_DATA_PAGES, /* data source/destination is a page array */
78 CEPH_MSG_DATA_PAGELIST, /* data source/destination is a pagelist */
79#ifdef CONFIG_BLOCK
80 CEPH_MSG_DATA_BIO, /* data source/destination is a bio list */
81#endif /* CONFIG_BLOCK */
82 CEPH_MSG_DATA_BVECS, /* data source/destination is a bio_vec array */
83};
84
85#ifdef CONFIG_BLOCK
86
87struct ceph_bio_iter {
88 struct bio *bio;
89 struct bvec_iter iter;
90};
91
92#define __ceph_bio_iter_advance_step(it, n, STEP) do { \
93 unsigned int __n = (n), __cur_n; \
94 \
95 while (__n) { \
96 BUG_ON(!(it)->iter.bi_size); \
97 __cur_n = min((it)->iter.bi_size, __n); \
98 (void)(STEP); \
99 bio_advance_iter((it)->bio, &(it)->iter, __cur_n); \
100 if (!(it)->iter.bi_size && (it)->bio->bi_next) { \
101 dout("__ceph_bio_iter_advance_step next bio\n"); \
102 (it)->bio = (it)->bio->bi_next; \
103 (it)->iter = (it)->bio->bi_iter; \
104 } \
105 __n -= __cur_n; \
106 } \
107} while (0)
108
109/*
110 * Advance @it by @n bytes.
111 */
112#define ceph_bio_iter_advance(it, n) \
113 __ceph_bio_iter_advance_step(it, n, 0)
114
115/*
116 * Advance @it by @n bytes, executing BVEC_STEP for each bio_vec.
117 */
118#define ceph_bio_iter_advance_step(it, n, BVEC_STEP) \
119 __ceph_bio_iter_advance_step(it, n, ({ \
120 struct bio_vec bv; \
121 struct bvec_iter __cur_iter; \
122 \
123 __cur_iter = (it)->iter; \
124 __cur_iter.bi_size = __cur_n; \
125 __bio_for_each_segment(bv, (it)->bio, __cur_iter, __cur_iter) \
126 (void)(BVEC_STEP); \
127 }))
128
129#endif /* CONFIG_BLOCK */
130
131struct ceph_bvec_iter {
132 struct bio_vec *bvecs;
133 struct bvec_iter iter;
134};
135
136#define __ceph_bvec_iter_advance_step(it, n, STEP) do { \
137 BUG_ON((n) > (it)->iter.bi_size); \
138 (void)(STEP); \
139 bvec_iter_advance((it)->bvecs, &(it)->iter, (n)); \
140} while (0)
141
142/*
143 * Advance @it by @n bytes.
144 */
145#define ceph_bvec_iter_advance(it, n) \
146 __ceph_bvec_iter_advance_step(it, n, 0)
147
148/*
149 * Advance @it by @n bytes, executing BVEC_STEP for each bio_vec.
150 */
151#define ceph_bvec_iter_advance_step(it, n, BVEC_STEP) \
152 __ceph_bvec_iter_advance_step(it, n, ({ \
153 struct bio_vec bv; \
154 struct bvec_iter __cur_iter; \
155 \
156 __cur_iter = (it)->iter; \
157 __cur_iter.bi_size = (n); \
158 for_each_bvec(bv, (it)->bvecs, __cur_iter, __cur_iter) \
159 (void)(BVEC_STEP); \
160 }))
161
162#define ceph_bvec_iter_shorten(it, n) do { \
163 BUG_ON((n) > (it)->iter.bi_size); \
164 (it)->iter.bi_size = (n); \
165} while (0)
166
167struct ceph_msg_data {
168 enum ceph_msg_data_type type;
169 union {
170#ifdef CONFIG_BLOCK
171 struct {
172 struct ceph_bio_iter bio_pos;
173 u32 bio_length;
174 };
175#endif /* CONFIG_BLOCK */
176 struct ceph_bvec_iter bvec_pos;
177 struct {
178 struct page **pages;
179 size_t length; /* total # bytes */
180 unsigned int alignment; /* first page */
181 bool own_pages;
182 };
183 struct ceph_pagelist *pagelist;
184 };
185};
186
187struct ceph_msg_data_cursor {
188 size_t total_resid; /* across all data items */
189
190 struct ceph_msg_data *data; /* current data item */
191 size_t resid; /* bytes not yet consumed */
192 bool last_piece; /* current is last piece */
193 bool need_crc; /* crc update needed */
194 union {
195#ifdef CONFIG_BLOCK
196 struct ceph_bio_iter bio_iter;
197#endif /* CONFIG_BLOCK */
198 struct bvec_iter bvec_iter;
199 struct { /* pages */
200 unsigned int page_offset; /* offset in page */
201 unsigned short page_index; /* index in array */
202 unsigned short page_count; /* pages in array */
203 };
204 struct { /* pagelist */
205 struct page *page; /* page from list */
206 size_t offset; /* bytes from list */
207 };
208 };
209};
210
211/*
212 * a single message. it contains a header (src, dest, message type, etc.),
213 * footer (crc values, mainly), a "front" message body, and possibly a
214 * data payload (stored in some number of pages).
215 */
216struct ceph_msg {
217 struct ceph_msg_header hdr; /* header */
218 union {
219 struct ceph_msg_footer footer; /* footer */
220 struct ceph_msg_footer_old old_footer; /* old format footer */
221 };
222 struct kvec front; /* unaligned blobs of message */
223 struct ceph_buffer *middle;
224
225 size_t data_length;
226 struct ceph_msg_data *data;
227 int num_data_items;
228 int max_data_items;
229 struct ceph_msg_data_cursor cursor;
230
231 struct ceph_connection *con;
232 struct list_head list_head; /* links for connection lists */
233
234 struct kref kref;
235 bool more_to_follow;
236 bool needs_out_seq;
237 int front_alloc_len;
238 unsigned long ack_stamp; /* tx: when we were acked */
239
240 struct ceph_msgpool *pool;
241};
242
243/* ceph connection fault delay defaults, for exponential backoff */
244#define BASE_DELAY_INTERVAL (HZ/2)
245#define MAX_DELAY_INTERVAL (5 * 60 * HZ)
246
247/*
248 * A single connection with another host.
249 *
250 * We maintain a queue of outgoing messages, and some session state to
251 * ensure that we can preserve the lossless, ordered delivery of
252 * messages in the case of a TCP disconnect.
253 */
254struct ceph_connection {
255 void *private;
256
257 const struct ceph_connection_operations *ops;
258
259 struct ceph_messenger *msgr;
260
261 atomic_t sock_state;
262 struct socket *sock;
263 struct ceph_entity_addr peer_addr; /* peer address */
264 struct ceph_entity_addr peer_addr_for_me;
265
266 unsigned long flags;
267 unsigned long state;
268 const char *error_msg; /* error message, if any */
269
270 struct ceph_entity_name peer_name; /* peer name */
271
272 u64 peer_features;
273 u32 connect_seq; /* identify the most recent connection
274 attempt for this connection, client */
275 u32 peer_global_seq; /* peer's global seq for this connection */
276
277 struct ceph_auth_handshake *auth;
278 int auth_retry; /* true if we need a newer authorizer */
279
280 struct mutex mutex;
281
282 /* out queue */
283 struct list_head out_queue;
284 struct list_head out_sent; /* sending or sent but unacked */
285 u64 out_seq; /* last message queued for send */
286
287 u64 in_seq, in_seq_acked; /* last message received, acked */
288
289 /* connection negotiation temps */
290 char in_banner[CEPH_BANNER_MAX_LEN];
291 struct ceph_msg_connect out_connect;
292 struct ceph_msg_connect_reply in_reply;
293 struct ceph_entity_addr actual_peer_addr;
294
295 /* message out temps */
296 struct ceph_msg_header out_hdr;
297 struct ceph_msg *out_msg; /* sending message (== tail of
298 out_sent) */
299 bool out_msg_done;
300
301 struct kvec out_kvec[8], /* sending header/footer data */
302 *out_kvec_cur;
303 int out_kvec_left; /* kvec's left in out_kvec */
304 int out_skip; /* skip this many bytes */
305 int out_kvec_bytes; /* total bytes left */
306 int out_more; /* there is more data after the kvecs */
307 __le64 out_temp_ack; /* for writing an ack */
308 struct ceph_timespec out_temp_keepalive2; /* for writing keepalive2
309 stamp */
310
311 /* message in temps */
312 struct ceph_msg_header in_hdr;
313 struct ceph_msg *in_msg;
314 u32 in_front_crc, in_middle_crc, in_data_crc; /* calculated crc */
315
316 char in_tag; /* protocol control byte */
317 int in_base_pos; /* bytes read */
318 __le64 in_temp_ack; /* for reading an ack */
319
320 struct timespec64 last_keepalive_ack; /* keepalive2 ack stamp */
321
322 struct delayed_work work; /* send|recv work */
323 unsigned long delay; /* current delay interval */
324};
325
326
327extern const char *ceph_pr_addr(const struct ceph_entity_addr *addr);
328
329extern int ceph_parse_ips(const char *c, const char *end,
330 struct ceph_entity_addr *addr,
331 int max_count, int *count);
332
333
334extern int ceph_msgr_init(void);
335extern void ceph_msgr_exit(void);
336extern void ceph_msgr_flush(void);
337
338extern void ceph_messenger_init(struct ceph_messenger *msgr,
339 struct ceph_entity_addr *myaddr);
340extern void ceph_messenger_fini(struct ceph_messenger *msgr);
341extern void ceph_messenger_reset_nonce(struct ceph_messenger *msgr);
342
343extern void ceph_con_init(struct ceph_connection *con, void *private,
344 const struct ceph_connection_operations *ops,
345 struct ceph_messenger *msgr);
346extern void ceph_con_open(struct ceph_connection *con,
347 __u8 entity_type, __u64 entity_num,
348 struct ceph_entity_addr *addr);
349extern bool ceph_con_opened(struct ceph_connection *con);
350extern void ceph_con_close(struct ceph_connection *con);
351extern void ceph_con_send(struct ceph_connection *con, struct ceph_msg *msg);
352
353extern void ceph_msg_revoke(struct ceph_msg *msg);
354extern void ceph_msg_revoke_incoming(struct ceph_msg *msg);
355
356extern void ceph_con_keepalive(struct ceph_connection *con);
357extern bool ceph_con_keepalive_expired(struct ceph_connection *con,
358 unsigned long interval);
359
360void ceph_msg_data_add_pages(struct ceph_msg *msg, struct page **pages,
361 size_t length, size_t alignment, bool own_pages);
362extern void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
363 struct ceph_pagelist *pagelist);
364#ifdef CONFIG_BLOCK
365void ceph_msg_data_add_bio(struct ceph_msg *msg, struct ceph_bio_iter *bio_pos,
366 u32 length);
367#endif /* CONFIG_BLOCK */
368void ceph_msg_data_add_bvecs(struct ceph_msg *msg,
369 struct ceph_bvec_iter *bvec_pos);
370
371struct ceph_msg *ceph_msg_new2(int type, int front_len, int max_data_items,
372 gfp_t flags, bool can_fail);
373extern struct ceph_msg *ceph_msg_new(int type, int front_len, gfp_t flags,
374 bool can_fail);
375
376extern struct ceph_msg *ceph_msg_get(struct ceph_msg *msg);
377extern void ceph_msg_put(struct ceph_msg *msg);
378
379extern void ceph_msg_dump(struct ceph_msg *msg);
380
381#endif