1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2001 Cluster File Systems, Inc. <braam@clusterfs.com>
6 * This file is part of Lustre, http://www.lustre.org.
8 * Lustre is free software; you can redistribute it and/or
9 * modify it under the terms of version 2 of the GNU General Public
10 * License as published by the Free Software Foundation.
12 * Lustre is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with Lustre; if not, write to the Free Software
19 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 * Lustre wire protocol definitions.
23 * We assume all nodes are either little-endian or big-endian, and we
24 * always send messages in the sender's native format. The receiver
25 * detects the message format by checking the 'magic' field of the message
26 * (see lustre_msg_swabbed() below).
28 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
29 * implemented either here, inline (trivial implementations) or in
30 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
31 * endian, in-place in the message buffer.
33 * A swabber takes a single pointer argument. The caller must already have
34 * verified that the length of the message buffer >= sizeof (type).
36 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
37 * may be defined that swabs just the variable part, after the caller has
38 * verified that the message buffer is large enough.
41 #ifndef _LUSTRE_IDL_H_
42 #define _LUSTRE_IDL_H_
45 # include <linux/ioctl.h>
46 # include <asm/types.h>
47 # include <linux/types.h>
48 # include <linux/list.h>
49 # include <linux/string.h> /* for strncpy, below */
50 # include <asm/byteorder.h>
53 # include <sys/types.h>
55 # include <asm/types.h>
58 # include <portals/list.h>
61 * this file contains all data structures used in Lustre interfaces:
62 * - obdo and obd_request records
63 * - mds_request records
75 static inline int obd_uuid_equals(struct obd_uuid *u1, struct obd_uuid *u2)
77 return strcmp(u1->uuid, u2->uuid) == 0;
80 static inline void obd_str2uuid(struct obd_uuid *uuid, char *tmp)
82 strncpy(uuid->uuid, tmp, sizeof(*uuid));
83 uuid->uuid[sizeof(*uuid) - 1] = '\0';
86 extern struct obd_uuid lctl_fake_uuid;
88 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
89 * FOO_REPLY_PORTAL is for incoming replies on the FOO
90 * FOO_BULK_PORTAL is for incoming bulk on the FOO
93 #define CONNMGR_REQUEST_PORTAL 1
94 #define CONNMGR_REPLY_PORTAL 2
95 //#define OSC_REQUEST_PORTAL 3
96 #define OSC_REPLY_PORTAL 4
97 //#define OSC_BULK_PORTAL 5
98 #define OST_REQUEST_PORTAL 6
99 //#define OST_REPLY_PORTAL 7
100 #define OST_BULK_PORTAL 8
101 //#define MDC_REQUEST_PORTAL 9
102 #define MDC_REPLY_PORTAL 10
103 //#define MDC_BULK_PORTAL 11
104 #define MDS_REQUEST_PORTAL 12
105 //#define MDS_REPLY_PORTAL 13
106 #define MDS_BULK_PORTAL 14
107 #define LDLM_CB_REQUEST_PORTAL 15
108 #define LDLM_CB_REPLY_PORTAL 16
109 #define LDLM_CANCEL_REQUEST_PORTAL 17
110 #define LDLM_CANCEL_REPLY_PORTAL 18
111 #define PTLBD_REQUEST_PORTAL 19
112 #define PTLBD_REPLY_PORTAL 20
113 #define PTLBD_BULK_PORTAL 21
114 #define MDS_SETATTR_PORTAL 22
115 #define MDS_READPAGE_PORTAL 23
116 #define MGMT_REQUEST_PORTAL 24
117 #define MGMT_REPLY_PORTAL 25
118 #define MGMT_CLI_REQUEST_PORTAL 26
119 #define MGMT_CLI_REPLY_PORTAL 27
124 #define SVC_RUNNING 8
125 #define SVC_STOPPING 16
126 #define SVC_STOPPED 32
128 #define LUSTRE_CONN_NEW 1
129 #define LUSTRE_CONN_CON 2
130 #define LUSTRE_CONN_NOTCONN 3
131 #define LUSTRE_CONN_RECOVER 4
132 #define LUSTRE_CONN_FULL 5
135 #define PTL_RPC_MSG_REQUEST 4711
136 #define PTL_RPC_MSG_ERR 4712
137 #define PTL_RPC_MSG_REPLY 4713
139 #define PTLRPC_MSG_MAGIC 0x0BD00BD0
140 #define PTLRPC_MSG_VERSION 0x00040002
142 struct lustre_handle {
145 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabe
147 /* we depend on this structure to be 8-byte aligned */
148 /* this type is only endian-adjusted in lustre_unpack_msg() */
150 struct lustre_handle handle;
156 __u64 last_committed;
164 static inline int lustre_msg_swabbed (struct lustre_msg *msg)
166 return (msg->magic == __swab32(PTLRPC_MSG_MAGIC));
169 /* Flags that are operation-specific go in the top 16 bits. */
170 #define MSG_OP_FLAG_MASK 0xffff0000
171 #define MSG_OP_FLAG_SHIFT 16
173 /* Flags that apply to all requests are in the bottom 16 bits */
174 #define MSG_GEN_FLAG_MASK 0x0000ffff
175 #define MSG_LAST_REPLAY 1
178 static inline int lustre_msg_get_flags(struct lustre_msg *msg)
180 return (msg->flags & MSG_GEN_FLAG_MASK);
183 static inline void lustre_msg_add_flags(struct lustre_msg *msg, int flags)
185 msg->flags |= MSG_GEN_FLAG_MASK & flags;
188 static inline void lustre_msg_set_flags(struct lustre_msg *msg, int flags)
190 msg->flags &= ~MSG_GEN_FLAG_MASK;
191 lustre_msg_add_flags(msg, flags);
194 static inline int lustre_msg_get_op_flags(struct lustre_msg *msg)
196 return (msg->flags >> MSG_OP_FLAG_SHIFT);
199 static inline void lustre_msg_add_op_flags(struct lustre_msg *msg, int flags)
201 msg->flags |= ((flags & MSG_GEN_FLAG_MASK) << MSG_OP_FLAG_SHIFT);
204 static inline void lustre_msg_set_op_flags(struct lustre_msg *msg, int flags)
206 msg->flags &= ~MSG_OP_FLAG_MASK;
207 lustre_msg_add_op_flags(msg, flags);
211 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
214 #define MSG_CONNECT_RECOVERING 0x1
215 #define MSG_CONNECT_RECONNECT 0x2
216 #define MSG_CONNECT_REPLAYABLE 0x4
217 #define MSG_CONNECT_PEER 0x8
220 * OST requests: OBDO & OBD request records
225 OST_REPLY = 0, /* reply ? */
245 #define OST_FIRST_OPC OST_REPLY
246 /* When adding OST RPC opcodes, please update
247 * LAST/FIRST macros used in ptlrpc/ptlrpc_internals.h */
249 typedef uint64_t obd_id;
250 typedef uint64_t obd_gr;
251 typedef uint64_t obd_time;
252 typedef uint64_t obd_size;
253 typedef uint64_t obd_off;
254 typedef uint64_t obd_blocks;
255 typedef uint32_t obd_blksize;
256 typedef uint32_t obd_mode;
257 typedef uint32_t obd_uid;
258 typedef uint32_t obd_gid;
259 typedef uint64_t obd_rdev;
260 typedef uint32_t obd_flag;
261 typedef uint32_t obd_count;
263 #define OBD_FL_INLINEDATA (0x00000001)
264 #define OBD_FL_OBDMDEXISTS (0x00000002)
266 #define OBD_INLINESZ 60
267 #define FD_OSTDATA_SIZE sizeof(struct obd_client_handle)
269 /* Note: 64-bit types are 64-bit aligned in structure */
277 obd_blocks o_blocks; /* brw: clients sent cached bytes */
278 obd_rdev o_rdev; /* brw: clients/servers sent grant */
279 obd_blksize o_blksize; /* optimal IO blocksize */
284 obd_count o_nlink; /* brw: checksum */
285 obd_count o_generation;
286 obd_flag o_valid; /* hot fields in this obdo */
289 char o_inline[OBD_INLINESZ];
292 extern void lustre_swab_obdo (struct obdo *o);
294 struct lov_object_id { /* per-child structure */
298 #define LOV_MAGIC 0x0BD00BD0
302 __u64 lmm_object_id; /* lov object id */
303 __u32 lmm_stripe_size; /* size of the stripe */
304 __u32 lmm_stripe_offset; /* starting stripe offset in lmm_objects */
305 __u16 lmm_stripe_count; /* number of stipes in use for this object */
306 __u16 lmm_ost_count; /* how many OST idx are in this LOV md */
307 struct lov_object_id lmm_objects[0];
308 } __attribute__((packed));
310 #define OBD_MD_FLALL (0xffffffff)
311 #define OBD_MD_FLID (0x00000001) /* object ID */
312 #define OBD_MD_FLATIME (0x00000002) /* access time */
313 #define OBD_MD_FLMTIME (0x00000004) /* data modification time */
314 #define OBD_MD_FLCTIME (0x00000008) /* change time */
315 #define OBD_MD_FLSIZE (0x00000010) /* size */
316 #define OBD_MD_FLBLOCKS (0x00000020) /* allocated blocks count */
317 #define OBD_MD_FLBLKSZ (0x00000040) /* block size */
318 #define OBD_MD_FLMODE (0x00000080) /* access bits (mode & ~S_IFMT) */
319 #define OBD_MD_FLTYPE (0x00000100) /* object type (mode & S_IFMT) */
320 #define OBD_MD_FLUID (0x00000200) /* user ID */
321 #define OBD_MD_FLGID (0x00000400) /* group ID */
322 #define OBD_MD_FLFLAGS (0x00000800) /* flags word */
323 #define OBD_MD_FLOBDFLG (0x00001000)
324 #define OBD_MD_FLNLINK (0x00002000) /* link count */
325 #define OBD_MD_FLGENER (0x00004000) /* generation number */
326 #define OBD_MD_FLINLINE (0x00008000) /* inline data */
327 #define OBD_MD_FLRDEV (0x00010000) /* device number */
328 #define OBD_MD_FLEASIZE (0x00020000) /* extended attribute data */
329 #define OBD_MD_LINKNAME (0x00040000) /* symbolic link target */
330 #define OBD_MD_FLHANDLE (0x00080000) /* file handle */
331 #define OBD_MD_FLCKSUM (0x00100000) /* bulk data checksum */
332 #define OBD_MD_FLQOS (0x00200000) /* quality of service stats */
333 #define OBD_MD_FLOSCOPQ (0x00400000) /* osc opaque data */
334 #define OBD_MD_FLCOOKIE (0x00800000) /* log cancellation cookie */
335 #define OBD_MD_FLNOTOBD (~(OBD_MD_FLOBDFLG | OBD_MD_FLBLOCKS | OBD_MD_LINKNAME|\
336 OBD_MD_FLEASIZE | OBD_MD_FLHANDLE | OBD_MD_FLCKSUM|\
337 OBD_MD_FLQOS | OBD_MD_FLOSCOPQ | OBD_MD_FLCOOKIE))
339 static inline struct lustre_handle *obdo_handle(struct obdo *oa)
341 return (struct lustre_handle *)oa->o_inline;
344 static inline struct llog_cookie *obdo_logcookie(struct obdo *oa)
346 return (struct llog_cookie *)(oa->o_inline +
347 sizeof(struct lustre_handle));
364 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
366 /* ost_body.data values for OST_BRW */
368 #define OBD_BRW_READ 0x01
369 #define OBD_BRW_WRITE 0x02
370 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
371 #define OBD_BRW_CREATE 0x04
372 #define OBD_BRW_SYNC 0x08
373 #define OBD_BRW_CHECK 0x10
374 #define OBD_BRW_FROM_GRANT 0x20
376 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
383 } __attribute__((packed));
385 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
387 /* multiple of 8 bytes => can array */
388 struct niobuf_remote {
392 } __attribute__((packed));
394 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
396 /* request structure for OST's */
398 #define OST_REQ_HAS_OA1 0x1
404 extern void lustre_swab_ost_body (struct ost_body *b);
413 MDS_GETATTR_NAME = 34,
426 #define MDS_FIRST_OPC MDS_GETATTR
427 /* When adding MDS RPC opcodes, please update
428 * LAST/FIRST macros used in ptlrpc/ptlrpc_internals.h */
434 #define REINT_SETATTR 1
435 #define REINT_CREATE 2
437 #define REINT_UNLINK 4
438 #define REINT_RENAME 5
442 /* the disposition of the intent outlines what was executed */
443 #define DISP_IT_EXECD 1
444 #define DISP_LOOKUP_EXECD (1 << 1)
445 #define DISP_LOOKUP_NEG (1 << 2)
446 #define DISP_LOOKUP_POS (1 << 3)
447 #define DISP_OPEN_CREATE (1 << 4)
448 #define DISP_OPEN_OPEN (1 << 5)
449 #define DISP_ENQ_COMPLETE (1<<6)
463 extern void lustre_swab_ll_fid (struct ll_fid *fid);
465 #define MDS_STATUS_CONN 1
466 #define MDS_STATUS_LOV 2
468 struct mds_status_req {
473 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
475 struct mds_fileh_body {
477 struct lustre_handle f_handle;
480 extern void lustre_swab_mds_fileh_body (struct mds_fileh_body *f);
485 struct lustre_handle handle;
486 __u64 size; /* Offset, in the case of MDS_READPAGE */
487 __u64 blocks; /* XID, in the case of MDS_READPAGE */
488 __u32 ino; /* make this a __u64 */
501 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
507 extern void lustre_swab_mds_body (struct mds_body *b);
509 /* This is probably redundant with OBD_MD_FLEASIZE, but we need an audit */
510 #define MDS_OPEN_HAS_EA 1 /* this open has an EA, for a delayed create*/
512 /* MDS update records */
514 //struct mds_update_record_hdr {
518 struct mds_rec_setattr {
525 struct ll_fid sa_fid;
537 /* Remove this once we declare it in include/linux/fs.h (v21 kernel patch?) */
538 #ifndef ATTR_CTIME_SET
539 #define ATTR_CTIME_SET 0x2000
542 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
544 struct mds_rec_create {
549 __u32 cr_flags; /* for use with open */
551 struct ll_fid cr_fid;
552 struct ll_fid cr_replayfid;
560 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
562 struct mds_rec_link {
569 struct ll_fid lk_fid1;
570 struct ll_fid lk_fid2;
573 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
575 struct mds_rec_unlink {
583 struct ll_fid ul_fid1;
584 struct ll_fid ul_fid2;
587 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
589 struct mds_rec_rename {
596 struct ll_fid rn_fid1;
597 struct ll_fid rn_fid2;
600 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
603 * LOV data structures
609 #define LOV_MAX_UUID_BUFFER_SIZE 8192
610 /* The size of the buffer the lov/mdc reserves for the
611 * array of UUIDs returned by the MDS. With the current
612 * protocol, this will limit the max number of OSTs per LOV */
615 __u32 ld_tgt_count; /* how many OBD's */
616 __u32 ld_active_tgt_count; /* how many active */
617 __u32 ld_default_stripe_count; /* how many objects are used */
618 __u64 ld_default_stripe_size; /* in bytes */
619 __u64 ld_default_stripe_offset; /* in bytes */
620 __u32 ld_pattern; /* RAID 0,1 etc */
621 struct obd_uuid ld_uuid;
624 extern void lustre_swab_lov_desc (struct lov_desc *ld);
629 /* opcodes -- MUST be distinct from OST/MDS opcodes */
634 LDLM_BL_CALLBACK = 104,
635 LDLM_CP_CALLBACK = 105,
638 #define LDLM_FIRST_OPC LDLM_ENQUEUE
639 /* When adding LDLM RPC opcodes, please update
640 * LAST/FIRST macros used in ptlrpc/ptlrpc_internals.h */
642 #define RES_NAME_SIZE 3
643 #define RES_VERSION_SIZE 4
646 __u64 name[RES_NAME_SIZE];
649 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
666 extern void lustre_swab_ldlm_extent (struct ldlm_extent *e);
672 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
674 /* Note this unaligned structure; as long as it's only used in ldlm_request
675 * below, we're probably fine. */
676 struct ldlm_resource_desc {
678 struct ldlm_res_id lr_name;
679 __u32 lr_version[RES_VERSION_SIZE];
682 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
684 struct ldlm_lock_desc {
685 struct ldlm_resource_desc l_resource;
686 ldlm_mode_t l_req_mode;
687 ldlm_mode_t l_granted_mode;
688 struct ldlm_extent l_extent;
689 __u32 l_version[RES_VERSION_SIZE];
692 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
694 struct ldlm_request {
696 struct ldlm_lock_desc lock_desc;
697 struct lustre_handle lock_handle1;
698 struct lustre_handle lock_handle2;
701 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
706 struct ldlm_res_id lock_resource_name;
707 struct lustre_handle lock_handle;
708 struct ldlm_extent lock_extent; /* XXX make this policy 1 &2 */
709 __u64 lock_policy_res1;
710 __u64 lock_policy_res2;
713 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
716 * ptlbd, portal block device requests
724 PTLBD_DISCONNECT = 205,
727 #define PTLBD_FIRST_OPC PTLBD_QUERY
728 /* When adding PTLBD RPC opcodes, please update
729 * LAST/FIRST macros used in ptlrpc/ptlrpc_internals.h */
739 extern void lustre_swab_ptlbd_op (struct ptlbd_op *op);
748 extern void lustre_swab_ptlbd_niob (struct ptlbd_niob *n);
755 extern void lustre_swab_ptlbd_rsp (struct ptlbd_rsp *r);
758 * Opcodes for management/monitoring node.
760 #define MGMT_CONNECT 250
761 #define MGMT_DISCONNECT 251
762 #define MGMT_EXCEPTION 252 /* node died, etc. */
765 * Opcodes for multiple servers.
769 #define OBD_LOG_CANCEL 401
770 #define OBD_LAST_OPC (OBD_LOG_CANCEL + 1)
771 #define OBD_FIRST_OPC OBD_PING
773 /* catalog of log objects */
775 /* Identifier for a single log object */
781 /* Log data record types - there is no specific reason that these need to
782 * be related to the RPC opcodes, but no reason not to (may be handy later?)
785 OST_CREATE_REC = 0x10600000 | (OST_CREATE << 8),
786 OST_ORPHAN_REC = 0x10600000 | (OST_DESTROY << 8),
787 MDS_UNLINK_REC = 0x10610000 | (MDS_REINT << 8) | REINT_UNLINK,
788 LLOG_CATALOG_MAGIC = 0x1062e67d,
789 LLOG_OBJECT_MAGIC = 0x10645539,
792 /* Log record header - stored in originating host endian order (use magic to
794 * Each record must start with this struct, end with a __u32 for the struct
795 * length, and be a multiple of 64 bits in size.
797 struct llog_trans_hdr {
802 struct llog_create_rec {
803 struct llog_trans_hdr lcr_hdr;
804 struct ll_fid lcr_fid;
808 } __attribute__((packed));
810 struct llog_orphan_rec {
811 struct llog_trans_hdr lor_hdr;
815 } __attribute__((packed));
817 struct llog_unlink_rec {
818 struct llog_trans_hdr lur_hdr;
822 } __attribute__((packed));
824 /* On-disk header structure of each log object - stored in creating host
825 * endian order, with the exception of the bitmap - stored in little endian
826 * order so that we can use ext2_{clear,set,test}_bit() for proper/optimized
827 * little-endian handling of bitmaps (which are otherwise a pain to handle).
829 #define LLOG_CHUNK_SIZE 4096
830 #define LLOG_HEADER_SIZE (96)
831 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
833 #define LLOG_MIN_REC_SIZE (16) /* round(struct llog_trans_hdr+end_len) */
835 struct llog_object_hdr {
836 struct llog_trans_hdr llh_hdr;
839 __u16 llh_bitmap_offset;
841 struct obd_uuid llh_tgtuuid;
843 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-17];
844 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
845 __u32 llh_hdr_end_len;
848 static inline int llog_log_swabbed(struct llog_object_hdr *hdr)
850 if (hdr->llh_hdr.lth_type == __swab32(LLOG_OBJECT_MAGIC))
852 if (hdr->llh_hdr.lth_type == LLOG_OBJECT_MAGIC)
857 /* log cookies are used to reference a specific log file and a record therein */
859 struct llog_logid lgc_lgl;