2 * Copyright (c) 2004-2007 The DragonFly Project. All rights reserved.
4 * This code is derived from software contributed to The DragonFly Project
5 * by Matthew Dillon <dillon@backplane.com>
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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
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.
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
34 * $DragonFly: src/sys/sys/syslink_msg.h,v 1.9 2007/06/17 21:31:07 dillon Exp $
37 * The syslink infrastructure implements an optimized RPC mechanism across a
38 * communications link. Endpoints, defined by a session sysid, are typically
39 * associated with system structures but do not have to be.
41 * This header file is primarily responsible for the formatting of message
42 * traffic over a syslink.
45 #ifndef _SYS_SYSLINK_MSG_H_
46 #define _SYS_SYSLINK_MSG_H_
49 #include <sys/types.h>
51 #ifndef _MACHINE_ATOMIC_H_
52 #include <machine/atomic.h>
55 typedef int32_t sl_auxdata_t; /* auxillary data element */
56 typedef u_int32_t sl_rlabel_t; /* reply label routing id */
57 typedef u_int16_t sl_proto_t; /* protocol control field */
58 typedef u_int16_t sl_cmd_t; /* command/status id */
59 typedef u_int16_t sl_reclen_t; /* item length */
61 #define SL_ALIGN 8 /* 8-byte alignment */
62 #define SL_ALIGNMASK (SL_ALIGN - 1)
65 * SYSLINK_ELM - structured data element.
67 * syslink_msg's have zero or more syslink_elm's arranged as an array.
68 * Each syslink_elm may represent opaque data or recursively structured
71 * SE_CMD field - identify RPC command (at the top level) or RPC data element
72 * in deeper recursions.
74 * Please note that while bits have individual meanings, command switches
75 * should universally compare all 16 bits against the command. This
76 * guarentees that commands will not be misinterpreted (e.g. reply vs
77 * command, or data which has not been endian converted).
79 * SE_CMDF_REPLY - is usually set in the top level syslink_elm embedded
80 * in syslink message replies as a safety in order to prevent a reply
81 * from being misinterpreted as a command.
83 * SE_CMDF_STRUCTURED - indicates that the payload is an array of
84 * structured syslink_elm's, otherwise the payload is considered to
87 * SE_CMDF_GLOBAL - indicates that the command is globally defined by the
88 * syslink standard and is not protocol-specific. Note that PADs
89 * are not global commands.
91 * SE_CMDF_UNTRANSLATED - indicates that the syslink_elm structure had
92 * to be translated into host endian format but any directly or
93 * indirectly represented opaque data has not been. This bit is used
94 * by the protocol layer to properly endian-translate protocol-specific
97 * SE_CMDF_ASIZE* - These 2 bits can encode the size for simple elments.
98 * The size is verified prior to delivery so command switches on simple
99 * elements need not check se_bytes. This also makes it easier for
100 * the protocol code to do endian conversions.
102 * SE_AUX field - auxillary data field (signed 32 bit integer)
104 * This field contains protocol and command/element specific data.
105 * This typically contains an error code in replies (at least in
109 sl_cmd_t se_cmd; /* syslink element command/status id */
110 sl_reclen_t se_bytes; /* unaligned record size */
111 sl_auxdata_t se_aux; /* auxillary data always present */
112 /* extended by data */
115 #define SE_CMDF_REPLY 0x8000 /* safety feature */
116 #define SE_CMDF_STRUCTURED 0x4000 /* payload is structured */
117 #define SE_CMDF_GLOBAL 0x2000 /* non-proto-specific global cmd */
118 #define SE_CMDF_UNTRANSLATED 0x1000 /* needs endian translation */
120 #define SE_CMDF_ASIZEMASK 0x0C00 /* auto-size mask */
121 #define SE_CMDF_ASIZEX 0x0000 /* N bytes of extended data */
122 #define SE_CMDF_ASIZE0 0x0400 /* 0 bytes of extended data */
123 #define SE_CMDF_ASIZE4 0x0800 /* 4 bytes of extended data */
124 #define SE_CMDF_ASIZE8 0x0C00 /* 8 bytes of extended data */
126 #define SE_CMD_MASK 0x03FF
128 #define SE_CMD_PAD 0x0000 /* always reserved to mean PAD */
131 * SYSLINK_MSG - Syslink transactional command or response
133 * This structure represents a syslink transactional command or response
134 * between two end-points identified by the session id. Either end may
135 * initiate a command independant of the other. A transaction consists of
136 * the sending of a command and the reception of a response.
138 * Multiple transactions in each direction (and both directions at once)
139 * may occur in parallel. The command/reply transaction space in one
140 * direction is independant of the command/reply transaction space in the
143 * SM_PROTO rppppppx-ppppppx
145 * r 0 = Command, 1 = Reply
147 * x Used to detect endian reversal. The protocol id is OR'd
148 * with 0x0100 on transmission. If we find bit 0 set to 1 on
149 * reception, endian translation must occur.
151 * - Reserved, must be 0
153 * p12 Encoded protocol number. Protocol 0 indicates PAD (r must
154 * be 0 as well). Protocols 0-63 are reserved and may only be
155 * used when officially recognized by the DragonFly project.
156 * 64-4095 are user defined.
158 * SM_BYTES bbbbbbbbbbbbbbbb
160 * b16 This is the size of the whole message, including headers
161 * but not including out-of-band DMA. All messages must
162 * be 8-byte aligned. Unlike syslink_elm structures, sm_bytes
163 * must be properly aligned.
165 * SM_RLABEL llllllllllllllllllllllllllllllll
167 * l32 This is a 32 bit reply label routing id. The format of
168 * this field is defined by the transport layer. The field
169 * is typically assigned in the command message as it passes
170 * through the transport layer and is retained verbatim in
173 * The most typical use of this field is as an aid to direct
174 * messages in a multi-threaded environment. For example,
175 * a kernel talking to a filesystem over a syslink might
176 * identify the thread originating the command in this field
177 * in order to allow the reply to be routed directly back to
180 * The field can also be used in crossbar switching meshes
181 * to identify both the originator and the target, but it
182 * should be noted that the verbatim requirement means the
183 * mesh must pick out the proper field based on the 'r'eply
188 * m64 This 64 bit message id combined with the mesh id and the
189 * 'r'eply bit (and also the direction of the message when
190 * operating over a full-duplex syslink) uniquely identifies
193 * The message id is typically set to the address of the
194 * syslink message or control structure used by the originator,
195 * or obtained from a 64 bit counter. This way the originator
196 * can guarentee uniqueness without actually having to track
197 * message id allocations.
201 * s64 This is a 64 bit session id key whos primary purpose is to
202 * validate a link and prevent improperly routed or stale
203 * messages from having an adverse effect on the cluster. The
204 * field is typically left 0 for intra-host links.
206 * SM_HEAD (structure)
208 * All syslink messages other then PAD messages must contain at least
209 * one whole syslink_elm. Elements are arranged in an array until
210 * the syslink message space is exhausted. Each element may represent
211 * opaque data or recursively structured data. Structured data consists
212 * of an array of 0 or more elements embedded in the parent element.
215 * ENDIAN TRANSLATION - endian translation occurs when a message is received
216 * with bit 0 set in sm_proto, indicating that the native endian mode of
217 * the sender is different from the native endian mode of the receiver.
218 * Endian translation is NOT specific to little or big endian formatting
219 * but instead occurs only when the two sides have different native endian
220 * modes. All fields are interpreted structurally. Only little and big
221 * endian formats are supported (i.e. simple byte reversal).
223 * Translation consists of reversing the byte ordering for each structural
224 * field. Any syslink_elm structures are recursively translated as well,
225 * but opaque data contained within is not. The SE_CMDF_UNTRANSLATED bit
226 * in each translated syslink_elm structure is flipped.
228 * Syslink routers and switches may or may not translate a syslink_msg (but
229 * they must still properly interpret the syslink_msg header as the
230 * message passes through). It is possible for a message to be translated
231 * multiple times while it transits the network so it is important when
232 * translation occurs that the SE_CMDF_UNTRANSLATED bit in the syslink_elm
233 * structures gets flipped rather then simply set.
236 sl_proto_t sm_proto; /* protocol id, endian, reply bit */
237 sl_reclen_t sm_bytes; /* unaligned size of message */
238 sl_rlabel_t sm_rlabel; /* reply label routing id */
239 /* minimum syslink_msg size is 8 bytes (special PAD) */
240 sysid_t sm_msgid; /* message id */
241 sysid_t sm_sessid; /* session id */
242 struct syslink_elm sm_head; /* structured data */
246 * Minimum sizes for syslink pads and syslink messages. Pads can be as
247 * small as 8 bytes and are 8-byte aligned. Syslink messages can be as
248 * small as 16 bytes and are 8-byte aligned.
250 #define SL_MIN_PAD_SIZE offsetof(struct syslink_msg, sm_msgid)
251 #define SL_MIN_MSG_SIZE sizeof(struct syslink_msg)
252 #define SL_MIN_ELM_SIZE sizeof(struct syslink_elm)
253 #define SL_MSG_ALIGN(bytes) (((bytes) + 7) & ~7)
256 * sm_proto field rppppppx-PPPPPPx encoded
257 * ----ppppppPPPPPP decoded
259 * Note: SMPROTO_ defines are in encoded form
261 #define SM_PROTO_REPLY 0x8000
262 #define SM_PROTO_ENDIAN_NORM 0x0100
263 #define SM_PROTO_ENDIAN_REV 0x0001
264 #define SM_PROTO_ENCODE(n) ((((n) << 1) & ~127) | (((n) << 3) & 0x7E00) \
265 | SM_PROTO_ENDIAN_NORM)
266 #define SM_PROTO_DECODE(n) ((((n) >> 1) & 63) | (((n) >> 3) & )) 0x0FC0) \
267 | SM_PROTO_ENDIAN_NORM)
270 * Reserved protocol encodings 0-63
272 #define SMPROTO_PAD SM_PROTO_ENCODE(0x0000)
275 * high level protocol encodings
277 #define SMPROTO_BSDVFS SM_PROTO_ENCODE(0x0040)
280 * Syslink messages may contain recursive components. The recursion depth
281 * allowed is limited to SL_MAXDEPTH.
283 * Syslink messages, NON-inclusive of any DMA buffers, are limited to
284 * SL_MAXSIZE bytes. DMA buffer limitations are not defined here but
285 * the expectation is that they can be fairly large.
287 #define SL_MAXDEPTH 10
288 #define SL_MAXSIZE 4096
293 #define SLMSG_SMALL 256
294 #define SLMSG_BIG SL_MAXSIZE
297 union syslink_small_msg {
298 struct syslink_msg msg;
299 char buf[SLMSG_SMALL];
302 union syslink_big_msg {
303 struct syslink_msg msg;
307 typedef struct syslink_msg *syslink_msg_t;
308 typedef struct syslink_elm *syslink_elm_t;