2 * Copyright (c) 2011-2012 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@dragonflybsd.org>
6 * by Venkatesh Srinivas <vsrinivas@dragonflybsd.org>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * 3. Neither the name of The DragonFly Project nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific, prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
28 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
30 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
31 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
32 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 /***************************************************************************
37 * LOW LEVEL MESSAGING *
38 ***************************************************************************
40 * hammer2_msg - A standalone copy of a message, typically referenced by
41 * or embedded in other structures, or used with I/O queues.
43 * These structures are strictly temporary, so they do not have to be
44 * particularly optimized for size. All possible message headers are
45 * directly embedded (any), and the message may contain a reference
46 * to allocated auxillary data. The structure is recycled quite often
49 * This structure is typically not used for storing persistent message
50 * state (see hammer2_persist for that).
53 struct hammer2_persist;
56 struct hammer2_iocom *iocom;
57 struct hammer2_persist *persist;
58 TAILQ_ENTRY(hammer2_msg) entry; /* queue */
59 char *aux_data; /* aux-data if any */
62 hammer2_any_t any; /* raw extended msg header */
65 typedef struct hammer2_msg hammer2_msg_t;
67 TAILQ_HEAD(hammer2_msg_queue, hammer2_msg);
68 typedef struct hammer2_msg_queue hammer2_msg_queue_t;
70 #define HAMMER2_MSGX_BSWAPPED 0x0001
73 * hammer2_ioq - An embedded component of hammer2_connect, holds state
74 * for the buffering and parsing of incoming and outgoing messages.
77 enum { HAMMER2_MSGQ_STATE_HEADER1,
78 HAMMER2_MSGQ_STATE_HEADER2,
79 HAMMER2_MSGQ_STATE_AUXDATA1,
80 HAMMER2_MSGQ_STATE_AUXDATA2,
81 HAMMER2_MSGQ_STATE_ERROR } state;
82 int fifo_beg; /* buffered data */
84 int hbytes; /* header size */
85 int abytes; /* aux_data size */
87 int seq; /* salt sequencer */
90 hammer2_msg_queue_t msgq;
93 typedef struct hammer2_ioq hammer2_ioq_t;
95 #define HAMMER2_IOQ_ERROR_SYNC 1 /* bad magic / out of sync */
96 #define HAMMER2_IOQ_ERROR_EOF 2 /* unexpected EOF */
97 #define HAMMER2_IOQ_ERROR_SOCK 3 /* read() error on socket */
98 #define HAMMER2_IOQ_ERROR_FIELD 4 /* invalid field */
99 #define HAMMER2_IOQ_ERROR_HCRC 5 /* core header crc bad */
100 #define HAMMER2_IOQ_ERROR_XCRC 6 /* ext header crc bad */
101 #define HAMMER2_IOQ_ERROR_ACRC 7 /* aux data crc bad */
102 #define HAMMER2_IOQ_ERROR_STATE 8 /* bad state */
104 #define HAMMER2_IOQ_MAXIOVEC 16
107 * hammer2_iocom - governs a messaging stream connection
109 struct hammer2_iocom {
110 hammer2_ioq_t ioq_rx;
111 hammer2_ioq_t ioq_tx;
112 hammer2_msg_queue_t freeq; /* free msgs hdr only */
113 hammer2_msg_queue_t freeq_aux; /* free msgs w/aux_data */
114 void (*recvmsg_callback)(struct hammer2_iocom *);
115 void (*sendmsg_callback)(struct hammer2_iocom *);
116 void (*altmsg_callback)(struct hammer2_iocom *);
117 int sock_fd; /* comm socket or pipe */
118 int alt_fd; /* thread signal, tty, etc */
120 char rxbuf[HAMMER2_MSGBUF_SIZE]; /* for ioq_rx only */
123 typedef struct hammer2_iocom hammer2_iocom_t;
125 #define HAMMER2_IOCOMF_EOF 0x00000001 /* EOF or ERROR on desc */
126 #define HAMMER2_IOCOMF_RREQ 0x00000002 /* request read-data event */
127 #define HAMMER2_IOCOMF_WREQ 0x00000004 /* request write-avail event */
128 #define HAMMER2_IOCOMF_WIDLE 0x00000008 /* request write-avail event */
129 #define HAMMER2_IOCOMF_SIGNAL 0x00000010
131 /***************************************************************************
132 * HIGH LEVEL MESSAGING *
133 ***************************************************************************
135 * Persistent state is stored via the hammer2_persist structure.
137 struct hammer2_persist {
138 uint32_t lcmd; /* recent command direction */
139 uint32_t lrep; /* recent reply direction */
142 typedef struct hammer2_persist hammer2_persist_t;
149 * The global registration structure consolidates information accumulated
150 * via the spanning tree algorithm and tells us which connection (link)
151 * is the best path to get to any given registration.
153 * glob_node - Splay entry for this registration in the global index
154 * of all registrations.
156 * glob_entry - tailq entry when this registration's best_span element
159 * span_list - Head of a simple list of spanning tree entries which
160 * we use to determine the best link.
162 * best_span - Which of the span structure on span_list is the best
165 * source_root - Splay tree root indexing all mesasges sent from this
166 * registration. The messages are indexed by
169 * target_root - Splay tree root indexing all messages being sent to
170 * this registration. The messages are indexed by
171 * {linkid,msgid}. XXX
174 * Whenever spanning tree data causes a registration's best_link field to
175 * change that registration is transmitted as spanning tree data to every
176 * active link. Note that pure clients to the cluster, of which there can
177 * be millions, typically do not transmit spanning tree data to each other.
179 * Each registration is assigned a unique linkid local to the node (another
180 * node might assign a different linkid to the same registration). This
181 * linkid must be persistent as long as messages are active and is used
182 * to identify the message source and target.
184 TAILQ_HEAD(hammer2_span_list, hammer2_span);
185 typedef struct hammer2_span_list hammer2_span_list_t;
188 SPLAY_ENTRY(hammer2_reg) glob_node; /* index of registrations */
189 TAILQ_ENTRY(hammer2_reg) glob_entry; /* when modified */
190 hammer2_span_list_t span_list; /* list of hammer2_span's */
191 hammer2_span_t *best_span; /* best span entry */
192 hammer2_pmsg_splay_head_t source_root; /* msgs sent from reg */
193 hammer2_pmsg_splay_head_t target_root; /* msgs sent to reg */
194 uuid_t pfs_id; /* key field */
195 uuid_t pfs_fsid; /* key field */
201 #define HAMMER2_PROTO_REGF_MODIFIED 0x0001
204 * Each link (connection) collects spanning tree data received via the
205 * link and stores it in these span structures.
207 struct hammer2_span {
208 TAILQ_ENTRY(hammer2_span) span_entry; /* from hammer2_reg */
209 SPLAY_ENTRY(hammer2_span) span_node; /* from hammer2_link */
211 hammer2_link_t *link;
216 * Most hammer2 messages represent transactions and have persistent state
217 * which must be recorded. Some messages, such as cache states and inode
218 * representations are very long-lasting transactions.
220 * Each node in the graph must keep track of the message state in order
221 * to perform the proper action when a connection is lost. To do this
222 * the message is indexed on the source and target (global) registration,
223 * and the actual span element the message was received on and transmitted
224 * to is recorded (allowing us to retrieve the physical links involved).
226 * The {source_reg, target_reg, msgid} uniquely identifies a message. Any
227 * streaming operations using the same msgid use the same rendezvous.
229 * It is important to note that recorded state must use the same physical
230 * link (and thus the same chain of links across the graph) as was 'forged'
231 * by the initial message for that msgid. If the source span a message is
232 * received on does not match the recorded source, or the recorded target
233 * is no longer routeable, the message will be returned or generate an ABORT
234 * with LINKFAIL as appropriate.
236 struct hammer2_pmsg {
237 SPLAY_ENTRY(hammer2_pmsg) source_reg;
238 SPLAY_ENTRY(hammer2_pmsg) target_reg;
239 hammer2_span_t *source;
240 hammer2_span_t *target;