hammer2 - cluster / libdmsg circuit work
[dragonfly.git] / lib / libdmsg / dmsg.h
CommitLineData
0c3a8cd0
MD
1/*
2 * Copyright (c) 2011-2012 The DragonFly Project. All rights reserved.
3 *
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@dragonflybsd.org>
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 *
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
16 * distribution.
17 * 3. Neither the name of The DragonFly Project nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific, prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35#include <sys/types.h>
36#include <sys/queue.h>
37#include <sys/tree.h>
38#include <sys/dmsg.h>
39
40#include <pthread.h>
41
42#if 0
43#include <openssl/rsa.h> /* public/private key functions */
44#include <openssl/pem.h> /* public/private key file load */
45#endif
46#include <openssl/err.h>
47#include <openssl/evp.h> /* aes_256_cbc functions */
48
49#define DMSG_DEFAULT_DIR "/etc/hammer2"
50#define DMSG_PATH_REMOTE DMSG_DEFAULT_DIR "/remote"
51
52#define DMSG_LISTEN_PORT 987
53
54/***************************************************************************
55 * CRYPTO HANDSHAKE *
56 ***************************************************************************
57 *
58 * The initial public-key exchange is implementing by transmitting a
59 * 512-byte buffer to the other side in a symmetrical fashion. This
60 * buffer contains the following:
61 *
62 * (1) A random session key. 512 bits is specified. We use aes_256_cbc()
63 * and initialize the key with the first 256 bits and the iv[] with
64 * the second. Note that the transmitted and received session
65 * keys are XOR'd together to create the session key used for
66 * communications (so even if the verifier is compromised the session
67 * will still be gobbly gook if the public key has not been completely
68 * broken).
69 *
70 * (2) A verifier to determine that the decode was successful. It encodes
71 * an XOR of each group of 4 bytes from the session key.
72 *
73 * (3) Additional configuration and additional random data.
74 *
75 * - The hammer2 message header magic for endian detect
76 *
77 * - The hammer2 protocol version. The two sides agree on the
78 * smaller of the two.
79 *
80 * - All unused fields (junk*) are filled with random data.
81 *
82 * This structure must be exactly 512 bytes and expects to use 256-byte
83 * RSA keys.
84 */
85struct dmsg_handshake {
86 char pad1[8]; /* 000 */
87 uint16_t magic; /* 008 DMSG_HDR_MAGIC for endian detect */
88 uint16_t version; /* 00A hammer2 protocol version */
89 uint32_t flags; /* 00C protocol extension flags */
90 uint8_t sess[64]; /* 010 512-bit session key */
91 uint8_t verf[16]; /* 050 verifier = ~sess */
92 char quickmsg[32]; /* 060 reason for connecting */
93 char junk080[128]; /* 080-0FF */
94 char pad2[8]; /* 100-107 */
95 char junk100[256-8]; /* 108-1FF */
96};
97
98typedef struct dmsg_handshake dmsg_handshake_t;
99
100
101#define DMSG_CRYPTO_CHUNK_SIZE DMSG_ALIGN
102#define DMSG_MAX_IV_SIZE 32
103
104#define DMSG_CRYPTO_GCM_IV_FIXED_SIZE 4
105#define DMSG_CRYPTO_GCM_IV_SIZE 12
106#define DMSG_CRYPTO_GCM_KEY_SIZE 32
107#define DMSG_CRYPTO_GCM_TAG_SIZE 16
108
109#define DMSG_CRYPTO_ALGO_GCM_IDX 0
110
111#define DMSG_CRYPTO_ALGO DMSG_CRYPTO_ALGO_GCM_IDX
112
0c3a8cd0
MD
113/***************************************************************************
114 * LOW LEVEL MESSAGING *
115 ***************************************************************************
116 *
117 * dmsg_msg - A standalone copy of a message, typically referenced by
118 * or embedded in other structures, or used with I/O queues.
119 *
120 * These structures are strictly temporary, so they do not have to be
121 * particularly optimized for size. All possible message headers are
122 * directly embedded (any), and the message may contain a reference
123 * to allocated auxillary data. The structure is recycled quite often
124 * by a connection.
0c3a8cd0
MD
125 */
126struct dmsg_iocom;
0d20ec8a 127struct dmsg_circuit;
0c3a8cd0 128struct dmsg_state;
0c3a8cd0
MD
129struct dmsg_msg;
130
131TAILQ_HEAD(dmsg_state_queue, dmsg_state);
132TAILQ_HEAD(dmsg_msg_queue, dmsg_msg);
133RB_HEAD(dmsg_state_tree, dmsg_state);
0d20ec8a 134RB_HEAD(dmsg_circuit_tree, dmsg_circuit);
0c3a8cd0
MD
135
136struct h2span_link;
137struct h2span_relay;
138struct h2span_conn;
139
0d20ec8a
MD
140struct dmsg_circuit {
141 RB_ENTRY(dmsg_circuit) rbnode;
142 uint64_t msgid;
143 struct dmsg_iocom *iocom;
144 struct dmsg_state_tree staterd_tree; /* active transactions */
145 struct dmsg_state_tree statewr_tree; /* active transactions */
146 struct dmsg_circuit *peer; /* (if circuit relay) */
147 struct dmsg_state *state; /* open VC transaction state */
148 struct dmsg_state *span_state; /* span, relay or link */
149 int is_relay; /* span is h2span_relay */
150 int refs;
151};
152
153/*
154 * The state structure is ref-counted. The iocom cannot go away while
155 * state structures are active. However, the related h2span_* linkages
156 * can be destroyed and NULL'd out if the state is terminated in both
157 * directions.
158 */
0c3a8cd0
MD
159struct dmsg_state {
160 RB_ENTRY(dmsg_state) rbnode; /* indexed by msgid */
161 struct dmsg_iocom *iocom;
0d20ec8a
MD
162 struct dmsg_circuit *circuit; /* associated circuit */
163 uint32_t icmd; /* command creating state */
0c3a8cd0
MD
164 uint32_t txcmd; /* mostly for CMDF flags */
165 uint32_t rxcmd; /* mostly for CMDF flags */
166 uint64_t msgid; /* {spanid,msgid} uniq */
167 int flags;
168 int error;
0d20ec8a
MD
169 int refs; /* prevent destruction */
170 struct dmsg_msg *msg; /* msg creating orig state */
0c3a8cd0
MD
171 void (*func)(struct dmsg_msg *);
172 union {
173 void *any;
174 struct h2span_link *link;
175 struct h2span_conn *conn;
176 struct h2span_relay *relay;
0d20ec8a 177 struct dmsg_circuit *circ;
0c3a8cd0
MD
178 } any;
179};
180
181#define DMSG_STATE_INSERTED 0x0001
182#define DMSG_STATE_DYNAMIC 0x0002
183#define DMSG_STATE_NODEID 0x0004 /* manages a node id */
184
0d20ec8a
MD
185/*
186 * This is the core in-memory representation of a message structure.
187 * The iocom represents the incoming or outgoing iocom. Various state
188 * pointers are calculated based on the message's raw source and target
189 * fields, and will ref the underlying state. Message headers are embedded
190 * while auxillary data is separately allocated.
191 */
0c3a8cd0
MD
192struct dmsg_msg {
193 TAILQ_ENTRY(dmsg_msg) qentry;
0d20ec8a
MD
194 struct dmsg_iocom *iocom; /* incoming/outgoing iocom */
195 struct dmsg_circuit *circuit; /* associated circuit */
196 struct dmsg_state *state; /* message state */
0c3a8cd0
MD
197 size_t hdr_size;
198 size_t aux_size;
199 char *aux_data;
200 dmsg_any_t any;
201};
202
0d20ec8a 203typedef struct dmsg_circuit dmsg_circuit_t;
0c3a8cd0
MD
204typedef struct dmsg_state dmsg_state_t;
205typedef struct dmsg_msg dmsg_msg_t;
206typedef struct dmsg_msg_queue dmsg_msg_queue_t;
207
208int dmsg_state_cmp(dmsg_state_t *state1, dmsg_state_t *state2);
209RB_PROTOTYPE(dmsg_state_tree, dmsg_state, rbnode, dmsg_state_cmp);
0d20ec8a
MD
210int dmsg_circuit_cmp(dmsg_circuit_t *circuit1, dmsg_circuit_t *circuit2);
211RB_PROTOTYPE(dmsg_circuit_tree, dmsg_circuit, rbnode, dmsg_circuit_cmp);
0c3a8cd0
MD
212
213/*
214 * dmsg_ioq - An embedded component of dmsg_conn, holds state
215 * for the buffering and parsing of incoming and outgoing messages.
216 *
217 * cdx - beg - processed buffer data, encrypted or decrypted
218 * end - cdn - unprocessed buffer data not yet encrypted or decrypted
219 */
220struct dmsg_ioq {
221 enum { DMSG_MSGQ_STATE_HEADER1,
222 DMSG_MSGQ_STATE_HEADER2,
223 DMSG_MSGQ_STATE_AUXDATA1,
224 DMSG_MSGQ_STATE_AUXDATA2,
225 DMSG_MSGQ_STATE_ERROR } state;
226 size_t fifo_beg; /* buffered data */
227 size_t fifo_cdx; /* cdx-beg processed */
228 size_t fifo_cdn; /* end-cdn unprocessed */
229 size_t fifo_end;
230 size_t hbytes; /* header size */
231 size_t abytes; /* aux_data size */
232 int error;
233 int seq; /* salt sequencer */
234 int msgcount;
235 EVP_CIPHER_CTX ctx;
236 char iv[DMSG_MAX_IV_SIZE]; /* encrypt or decrypt iv[] */
237 dmsg_msg_t *msg;
238 dmsg_msg_queue_t msgq;
239 char buf[DMSG_BUF_SIZE]; /* staging buffer */
240};
241
242typedef struct dmsg_ioq dmsg_ioq_t;
243
244#define DMSG_IOQ_ERROR_SYNC 1 /* bad magic / out of sync */
245#define DMSG_IOQ_ERROR_EOF 2 /* unexpected EOF */
246#define DMSG_IOQ_ERROR_SOCK 3 /* read() error on socket */
247#define DMSG_IOQ_ERROR_FIELD 4 /* invalid field */
248#define DMSG_IOQ_ERROR_HCRC 5 /* core header crc bad */
249#define DMSG_IOQ_ERROR_XCRC 6 /* ext header crc bad */
250#define DMSG_IOQ_ERROR_ACRC 7 /* aux data crc bad */
251#define DMSG_IOQ_ERROR_STATE 8 /* bad state */
252#define DMSG_IOQ_ERROR_NOPEER 9 /* bad socket peer */
253#define DMSG_IOQ_ERROR_NORKEY 10 /* no remote keyfile found */
254#define DMSG_IOQ_ERROR_NOLKEY 11 /* no local keyfile found */
255#define DMSG_IOQ_ERROR_KEYXCHGFAIL 12 /* key exchange failed */
256#define DMSG_IOQ_ERROR_KEYFMT 13 /* key file format problem */
257#define DMSG_IOQ_ERROR_BADURANDOM 14 /* /dev/urandom is bad */
258#define DMSG_IOQ_ERROR_MSGSEQ 15 /* message sequence error */
259#define DMSG_IOQ_ERROR_EALREADY 16 /* ignore this message */
260#define DMSG_IOQ_ERROR_TRANS 17 /* state transaction issue */
261#define DMSG_IOQ_ERROR_IVWRAP 18 /* IVs exhaused */
262#define DMSG_IOQ_ERROR_MACFAIL 19 /* MAC of encr alg failed */
263#define DMSG_IOQ_ERROR_ALGO 20 /* Misc. encr alg error */
0d20ec8a
MD
264#define DMSG_IOQ_ERROR_ROUTED 21 /* ignore routed message */
265#define DMSG_IOQ_ERROR_BAD_CIRCUIT 22 /* unconfigured circuit */
266#define DMSG_IOQ_ERROR_UNUSED23 23
267#define DMSG_IOQ_ERROR_ASSYM 24 /* Assymetric path */
0c3a8cd0
MD
268
269#define DMSG_IOQ_MAXIOVEC 16
270
0c3a8cd0
MD
271/*
272 * dmsg_iocom - governs a messaging stream connection
273 */
274struct dmsg_iocom {
275 dmsg_ioq_t ioq_rx;
276 dmsg_ioq_t ioq_tx;
277 dmsg_msg_queue_t freeq; /* free msgs hdr only */
278 dmsg_msg_queue_t freeq_aux; /* free msgs w/aux_data */
279 int sock_fd; /* comm socket or pipe */
280 int alt_fd; /* thread signal, tty, etc */
281 int wakeupfds[2]; /* pipe wakes up iocom thread */
282 int flags;
283 int rxmisc;
284 int txmisc;
0d20ec8a
MD
285 void (*signal_callback)(struct dmsg_iocom *);
286 void (*rcvmsg_callback)(struct dmsg_msg *);
287 void (*altmsg_callback)(struct dmsg_iocom *);
288 void (*dbgmsg_callback)(dmsg_msg_t *msg);
289 struct dmsg_circuit_tree circuit_tree; /* active circuits */
290 struct dmsg_circuit circuit0; /* embedded circuit0 */
291 dmsg_msg_queue_t txmsgq; /* tx msgq from remote */
292 struct h2span_conn *conn; /* if LNK_CONN active */
0c3a8cd0
MD
293 pthread_mutex_t mtx; /* mutex for state*tree/rmsgq */
294};
295
296typedef struct dmsg_iocom dmsg_iocom_t;
297
298#define DMSG_IOCOMF_EOF 0x00000001 /* EOF or ERROR on desc */
299#define DMSG_IOCOMF_RREQ 0x00000002 /* request read-data event */
300#define DMSG_IOCOMF_WREQ 0x00000004 /* request write-avail event */
301#define DMSG_IOCOMF_RWORK 0x00000008 /* immediate work pending */
302#define DMSG_IOCOMF_WWORK 0x00000010 /* immediate work pending */
303#define DMSG_IOCOMF_PWORK 0x00000020 /* immediate work pending */
304#define DMSG_IOCOMF_ARWORK 0x00000040 /* immediate work pending */
305#define DMSG_IOCOMF_AWWORK 0x00000080 /* immediate work pending */
306#define DMSG_IOCOMF_SWORK 0x00000100 /* immediate work pending */
307#define DMSG_IOCOMF_CRYPTED 0x00000200 /* encrypt enabled */
308
309/*
310 * Crypto algorithm table and related typedefs.
311 */
0c3a8cd0
MD
312typedef int (*algo_init_fn)(dmsg_ioq_t *, char *, int, char *, int, int);
313typedef int (*algo_enc_fn)(dmsg_ioq_t *, char *, char *, int, int *);
314typedef int (*algo_dec_fn)(dmsg_ioq_t *, char *, char *, int, int *);
315
316struct crypto_algo {
317 const char *name;
318 int keylen;
319 int taglen;
320 algo_init_fn init;
321 algo_enc_fn enc_chunk;
322 algo_dec_fn dec_chunk;
323};
324
11f7caf4
MD
325/*
326 * Master service thread info
327 */
328struct dmsg_master_service_info {
329 int fd;
330 int detachme;
185ace93 331 void *handle;
11f7caf4 332 void (*dbgmsg_callback)(dmsg_msg_t *msg);
185ace93 333 void (*exit_callback)(void *handle);
11f7caf4
MD
334};
335
336typedef struct dmsg_master_service_info dmsg_master_service_info_t;
337
0d20ec8a
MD
338/*
339 * node callbacks
340 */
341#define DMSG_NODEOP_ADD 1
342#define DMSG_NODEOP_DEL 2
343
344extern void (*dmsg_node_handler)(void **opaquep, struct dmsg_msg *msg, int op);
345
11f7caf4 346
0c3a8cd0
MD
347/*
348 * icrc
349 */
350uint32_t dmsg_icrc32(const void *buf, size_t size);
351uint32_t dmsg_icrc32c(const void *buf, size_t size, uint32_t crc);
352
353/*
354 * debug
355 */
356const char *dmsg_basecmd_str(uint32_t cmd);
357const char *dmsg_msg_str(dmsg_msg_t *msg);
358
359/*
360 * subs
361 */
362void *dmsg_alloc(size_t bytes);
363void dmsg_free(void *ptr);
364const char *dmsg_uuid_to_str(uuid_t *uuid, char **strp);
0d20ec8a
MD
365const char *dmsg_peer_type_to_str(uint8_t type);
366const char *dmsg_pfs_type_to_str(uint8_t type);
0c3a8cd0
MD
367int dmsg_connect(const char *hostname);
368
369/*
370 * Msg support functions
371 */
372void dmsg_bswap_head(dmsg_hdr_t *head);
373void dmsg_ioq_init(dmsg_iocom_t *iocom, dmsg_ioq_t *ioq);
374void dmsg_ioq_done(dmsg_iocom_t *iocom, dmsg_ioq_t *ioq);
375void dmsg_iocom_init(dmsg_iocom_t *iocom, int sock_fd, int alt_fd,
0d20ec8a 376 void (*state_func)(dmsg_iocom_t *),
0c3a8cd0 377 void (*rcvmsg_func)(dmsg_msg_t *),
11f7caf4 378 void (*dbgmsg_func)(dmsg_msg_t *),
0c3a8cd0 379 void (*altmsg_func)(dmsg_iocom_t *));
0d20ec8a
MD
380void dmsg_iocom_restate(dmsg_iocom_t *iocom,
381 void (*state_func)(dmsg_iocom_t *),
0c3a8cd0
MD
382 void (*rcvmsg_func)(dmsg_msg_t *),
383 void (*altmsg_func)(dmsg_iocom_t *));
0d20ec8a 384void dmsg_iocom_signal(dmsg_iocom_t *iocom);
0c3a8cd0 385void dmsg_iocom_done(dmsg_iocom_t *iocom);
0d20ec8a
MD
386void dmsg_circuit_init(dmsg_iocom_t *iocom, dmsg_circuit_t *circuit);
387dmsg_msg_t *dmsg_msg_alloc(dmsg_circuit_t *circuit,
0c3a8cd0
MD
388 size_t aux_size, uint32_t cmd,
389 void (*func)(dmsg_msg_t *), void *data);
390void dmsg_msg_reply(dmsg_msg_t *msg, uint32_t error);
391void dmsg_msg_result(dmsg_msg_t *msg, uint32_t error);
392void dmsg_state_reply(dmsg_state_t *state, uint32_t error);
0d20ec8a 393void dmsg_state_result(dmsg_state_t *state, uint32_t error);
0c3a8cd0
MD
394
395void dmsg_msg_free(dmsg_msg_t *msg);
396
397void dmsg_iocom_core(dmsg_iocom_t *iocom);
398dmsg_msg_t *dmsg_ioq_read(dmsg_iocom_t *iocom);
399void dmsg_msg_write(dmsg_msg_t *msg);
400
401void dmsg_iocom_drain(dmsg_iocom_t *iocom);
402void dmsg_iocom_flush1(dmsg_iocom_t *iocom);
403void dmsg_iocom_flush2(dmsg_iocom_t *iocom);
404
405void dmsg_state_cleanuprx(dmsg_iocom_t *iocom, dmsg_msg_t *msg);
406void dmsg_state_free(dmsg_state_t *state);
0d20ec8a 407void dmsg_circuit_drop(dmsg_circuit_t *circuit);
0c3a8cd0 408
0d20ec8a 409int dmsg_circuit_relay(dmsg_msg_t *msg);
0c3a8cd0
MD
410
411/*
412 * Msg protocol functions
413 */
0d20ec8a 414void dmsg_msg_lnk_signal(dmsg_iocom_t *iocom);
0c3a8cd0
MD
415void dmsg_msg_lnk(dmsg_msg_t *msg);
416void dmsg_msg_dbg(dmsg_msg_t *msg);
0d20ec8a 417void dmsg_shell_tree(dmsg_circuit_t *circuit, char *cmdbuf __unused);
0c3a8cd0
MD
418
419/*
420 * Crypto functions
421 */
422void dmsg_crypto_setup(void);
423void dmsg_crypto_negotiate(dmsg_iocom_t *iocom);
424void dmsg_crypto_decrypt(dmsg_iocom_t *iocom, dmsg_ioq_t *ioq);
425int dmsg_crypto_encrypt(dmsg_iocom_t *iocom, dmsg_ioq_t *ioq,
426 struct iovec *iov, int n, size_t *nactp);
427
428/*
429 * Service daemon functions
430 */
431void *dmsg_master_service(void *data);
0d20ec8a 432void dmsg_circuit_printf(dmsg_circuit_t *circuit, const char *ctl, ...)
f7eac9d2 433 __printflike(2, 3);
0c3a8cd0
MD
434
435extern int DMsgDebugOpt;