1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * This file is part of Lustre, http://www.lustre.org
6 * Lustre wire protocol definitions.
9 * We assume all nodes are either little-endian or big-endian, and we
10 * always send messages in the sender's native format. The receiver
11 * detects the message format by checking the 'magic' field of the message
12 * (see lustre_msg_swabbed() below).
13 * ALL structs passing over the wire should be declared here. Structs
14 * that are used in interfaces with userspace should go in lustre_user.h.
16 * All structs being declared here should be built from simple fixed-size
17 * types (__u8, __u16, __u32, __u64) or be built from other types or
18 * structs also declared in this file. Similarly, all flags and magic
19 * values in those structs should also be declared here. This ensures
20 * that the Lustre wire protocol is not influenced by external dependencies.
22 * The only other acceptable items in this file are VERY SIMPLE accessor
23 * functions to avoid callers grubbing inside the structures, and the
24 * prototypes of the swabber functions for each struct. Nothing that
25 * depends on external functions or definitions should be in here.
27 * Structs must be properly aligned to put 64-bit values on an 8-byte
28 * boundary. Any structs being added here must also be added to
29 * utils/wirecheck.c and "make newwiretest" run to regenerate the
30 * utils/wiretest.c sources. This allows us to verify that wire structs
31 * have the proper alignment/size on all architectures.
33 * DO NOT CHANGE any of the structs, flags, values declared here and used
34 * in released Lustre versions. Some structs may have padding fields that
35 * can be used. Some structs might allow addition at the end (verify this
36 * in the code to ensure that new/old clients that see this larger struct
37 * do not fail, otherwise you need to implement protocol compatibility).
39 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
40 * implemented either here, inline (trivial implementations) or in
41 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
42 * endian, in-place in the message buffer.
44 * A swabber takes a single pointer argument. The caller must already have
45 * verified that the length of the message buffer >= sizeof (type).
47 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
48 * may be defined that swabs just the variable part, after the caller has
49 * verified that the message buffer is large enough.
52 #ifndef _LUSTRE_IDL_H_
53 #define _LUSTRE_IDL_H_
55 #include <libcfs/kp30.h>
57 #if defined(__linux__)
58 #include <linux/lustre_types.h>
59 #elif defined(__APPLE__)
60 #include <darwin/lustre_types.h>
61 #elif defined(__WINNT__)
62 #include <winnt/lustre_types.h>
64 #error Unsupported operating system.
67 /* Defn's shared with user-space. */
68 #include <lustre/lustre_user.h>
73 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
74 * FOO_REPLY_PORTAL is for incoming replies on the FOO
75 * FOO_BULK_PORTAL is for incoming bulk on the FOO
78 #define CONNMGR_REQUEST_PORTAL 1
79 #define CONNMGR_REPLY_PORTAL 2
80 //#define OSC_REQUEST_PORTAL 3
81 #define OSC_REPLY_PORTAL 4
82 //#define OSC_BULK_PORTAL 5
83 #define OST_IO_PORTAL 6
84 #define OST_CREATE_PORTAL 7
85 #define OST_BULK_PORTAL 8
86 //#define MDC_REQUEST_PORTAL 9
87 #define MDC_REPLY_PORTAL 10
88 //#define MDC_BULK_PORTAL 11
89 #define MDS_REQUEST_PORTAL 12
90 //#define MDS_REPLY_PORTAL 13
91 #define MDS_BULK_PORTAL 14
92 #define LDLM_CB_REQUEST_PORTAL 15
93 #define LDLM_CB_REPLY_PORTAL 16
94 #define LDLM_CANCEL_REQUEST_PORTAL 17
95 #define LDLM_CANCEL_REPLY_PORTAL 18
96 //#define PTLBD_REQUEST_PORTAL 19
97 //#define PTLBD_REPLY_PORTAL 20
98 //#define PTLBD_BULK_PORTAL 21
99 #define MDS_SETATTR_PORTAL 22
100 #define MDS_READPAGE_PORTAL 23
101 #define MDS_MDS_PORTAL 24
103 #define MGC_REPLY_PORTAL 25
104 #define MGS_REQUEST_PORTAL 26
105 #define MGS_REPLY_PORTAL 27
106 #define OST_REQUEST_PORTAL 28
107 #define FLD_REQUEST_PORTAL 29
108 #define SEQ_METADATA_PORTAL 30
109 #define SEQ_DATA_PORTAL 31
110 #define SEQ_CONTROLLER_PORTAL 32
115 #define SVC_RUNNING 8
116 #define SVC_STOPPING 16
117 #define SVC_STOPPED 32
120 #define PTL_RPC_MSG_REQUEST 4711
121 #define PTL_RPC_MSG_ERR 4712
122 #define PTL_RPC_MSG_REPLY 4713
124 /* DON'T use swabbed values of MAGIC as magic! */
125 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
126 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
128 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
129 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
131 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
133 #define PTLRPC_MSG_VERSION 0x00000003
134 #define LUSTRE_VERSION_MASK 0xffff0000
135 #define LUSTRE_OBD_VERSION 0x00010000
136 #define LUSTRE_MDS_VERSION 0x00020000
137 #define LUSTRE_OST_VERSION 0x00030000
138 #define LUSTRE_DLM_VERSION 0x00040000
139 #define LUSTRE_LOG_VERSION 0x00050000
140 #define LUSTRE_MGS_VERSION 0x00060000
142 typedef __u64 mdsno_t;
143 typedef __u64 seqno_t;
150 static inline __u64 range_space(struct lu_range *r)
152 return r->lr_end - r->lr_start;
155 static inline void range_zero(struct lu_range *r)
157 r->lr_start = r->lr_end = 0;
160 static inline int range_within(struct lu_range *r,
163 return s >= r->lr_start && s < r->lr_end;
166 static inline void range_alloc(struct lu_range *r,
170 r->lr_start = s->lr_start;
171 r->lr_end = s->lr_start + w;
175 static inline int range_is_sane(struct lu_range *r)
177 return (r->lr_end >= r->lr_start);
180 static inline int range_is_zero(struct lu_range *r)
182 return (r->lr_start == 0 && r->lr_end == 0);
185 static inline int range_is_exhausted(struct lu_range *r)
187 return range_space(r) == 0;
190 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x]"
192 #define PRANGE(range) \
197 __u64 f_seq; /* holds fid sequence. Lustre should support 2 ^ 64
198 * objects, thus even if one sequence has one object we
199 * reach this value. */
200 __u32 f_oid; /* fid number within its sequence. */
201 __u32 f_ver; /* holds fid version. */
208 LUSTRE_ROOT_FID_SEQ = 1ULL, /* XXX: should go into mkfs. */
210 /* initial fid id value */
211 LUSTRE_FID_INIT_OID = 1UL
214 /* get object sequence */
215 static inline __u64 fid_seq(const struct lu_fid *fid)
221 static inline __u32 fid_oid(const struct lu_fid *fid)
226 /* get object version */
227 static inline __u32 fid_ver(const struct lu_fid *fid)
232 static inline int fid_seq_is_sane(__u64 seq)
237 static inline void fid_zero(struct lu_fid *fid)
239 memset(fid, 0, sizeof(*fid));
242 static inline int fid_is_igif(const struct lu_fid *fid)
244 return fid_seq(fid) == LUSTRE_ROOT_FID_SEQ;
247 #define DFID "[0x%16.16"LPF64"x/0x%8.8x:0x%8.8x]"
255 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
257 /* check that all fields are converted */
258 CLASSERT(sizeof *src ==
259 sizeof fid_seq(src) +
260 sizeof fid_oid(src) + sizeof fid_ver(src));
261 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
262 dst->f_seq = cpu_to_be64(fid_seq(src));
263 dst->f_oid = cpu_to_be32(fid_oid(src));
264 dst->f_ver = cpu_to_be32(fid_ver(src));
267 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
269 /* check that all fields are converted */
270 CLASSERT(sizeof *src ==
271 sizeof fid_seq(src) +
272 sizeof fid_oid(src) + sizeof fid_ver(src));
273 dst->f_seq = be64_to_cpu(fid_seq(src));
274 dst->f_oid = be32_to_cpu(fid_oid(src));
275 dst->f_ver = be32_to_cpu(fid_ver(src));
276 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
280 * Storage representation for fids.
282 * Variable size, first byte contains the length of the whole record.
287 char fp_area[sizeof(struct lu_fid)];
290 void fid_pack(struct lu_fid_pack *pack, const struct lu_fid *fid,
291 struct lu_fid *befider);
292 void fid_unpack(const struct lu_fid_pack *pack, struct lu_fid *fid);
297 static inline int fid_is_sane(const struct lu_fid *fid)
301 ((fid_seq_is_sane(fid_seq(fid)) && fid_oid(fid) != 0
302 && fid_ver(fid) == 0) ||
306 static inline int fid_is_zero(const struct lu_fid *fid)
308 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
311 extern void lustre_swab_lu_fid(struct lu_fid *fid);
312 extern void lustre_swab_lu_range(struct lu_range *range);
314 static inline int lu_fid_eq(const struct lu_fid *f0,
315 const struct lu_fid *f1)
317 /* Check that there is no alignment padding. */
318 CLASSERT(sizeof *f0 ==
319 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
320 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
321 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
322 return memcmp(f0, f1, sizeof *f0) == 0;
326 * Layout of readdir pages, as transmitted on wire.
329 struct lu_fid lde_fid;
337 __u32 ldp_hash_start;
341 struct lu_dirent ldp_entries[0];
344 enum lu_dirpage_flags {
348 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
350 if (le16_to_cpu(dp->ldp_flags) & LDF_EMPTY)
353 return dp->ldp_entries;
356 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
358 struct lu_dirent *next;
360 if (le16_to_cpu(ent->lde_reclen) != 0)
361 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
368 static inline int lu_dirent_size(struct lu_dirent *ent)
370 if (le16_to_cpu(ent->lde_reclen) == 0) {
371 return (sizeof(*ent) +
372 le16_to_cpu(ent->lde_namelen) + 3) & ~3;
374 return le16_to_cpu(ent->lde_reclen);
377 #define DIR_END_OFF 0xfffffffeUL
379 struct lustre_handle {
382 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
384 static inline int lustre_handle_is_used(struct lustre_handle *lh)
386 return lh->cookie != 0ull;
389 static inline int lustre_handle_equal(struct lustre_handle *lh1,
390 struct lustre_handle *lh2)
392 return lh1->cookie == lh2->cookie;
395 static inline void lustre_handle_copy(struct lustre_handle *tgt,
396 struct lustre_handle *src)
398 tgt->cookie = src->cookie;
401 /* we depend on this structure to be 8-byte aligned */
402 /* this type is only endian-adjusted in lustre_unpack_msg() */
403 struct lustre_msg_v1 {
404 struct lustre_handle lm_handle;
410 __u64 lm_last_committed;
419 #define lustre_msg lustre_msg_v2
420 /* we depend on this structure to be 8-byte aligned */
421 /* this type is only endian-adjusted in lustre_unpack_msg() */
422 struct lustre_msg_v2 {
434 /* without gss, ptlrpc_body is put at the first buffer. */
436 struct lustre_handle pb_handle;
443 __u64 pb_last_committed;
455 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
457 /* message body offset for lustre_msg_v2 */
458 /* ptlrpc body offset in all request/reply messages */
459 #define MSG_PTLRPC_BODY_OFF 0
461 /* normal request/reply message record offset */
462 #define REQ_REC_OFF 1
463 #define REPLY_REC_OFF 1
465 /* ldlm request message body offset */
466 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
467 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
469 /* ldlm intent lock message body offset */
470 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
471 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
473 /* ldlm reply message body offset */
474 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
475 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
477 /* Flags that are operation-specific go in the top 16 bits. */
478 #define MSG_OP_FLAG_MASK 0xffff0000
479 #define MSG_OP_FLAG_SHIFT 16
481 /* Flags that apply to all requests are in the bottom 16 bits */
482 #define MSG_GEN_FLAG_MASK 0x0000ffff
483 #define MSG_LAST_REPLAY 1
486 #define MSG_REQ_REPLAY_DONE 8
487 #define MSG_LOCK_REPLAY_DONE 16
490 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
493 #define MSG_CONNECT_RECOVERING 0x1
494 #define MSG_CONNECT_RECONNECT 0x2
495 #define MSG_CONNECT_REPLAYABLE 0x4
496 //#define MSG_CONNECT_PEER 0x8
497 #define MSG_CONNECT_LIBCLIENT 0x10
498 #define MSG_CONNECT_INITIAL 0x20
499 #define MSG_CONNECT_ASYNC 0x40
500 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
501 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
504 #define OBD_CONNECT_RDONLY 0x00000001ULL /* client allowed read-only access */
505 #define OBD_CONNECT_INDEX 0x00000002ULL /* connect to specific LOV idx */
506 #define OBD_CONNECT_GRANT 0x00000008ULL /* OSC acquires grant at connect */
507 #define OBD_CONNECT_SRVLOCK 0x00000010ULL /* server takes locks for client */
508 #define OBD_CONNECT_VERSION 0x00000020ULL /* Server supports versions in ocd */
509 #define OBD_CONNECT_REQPORTAL 0x00000040ULL /* Separate portal for non-IO reqs */
510 #define OBD_CONNECT_ACL 0x00000080ULL /* client uses access control lists */
511 #define OBD_CONNECT_XATTR 0x00000100ULL /* client using extended attributes*/
512 #define OBD_CONNECT_TRUNCLOCK 0x00000400ULL /* locks on server for punch b=9528 */
513 #define OBD_CONNECT_IBITS 0x00001000ULL /* support for inodebits locks */
514 #define OBD_CONNECT_JOIN 0x00002000ULL /* files can be concatenated */
515 #define OBD_CONNECT_ATTRFID 0x00004000ULL /* Server supports GetAttr By Fid */
516 #define OBD_CONNECT_NODEVOH 0x00008000ULL /* No open handle for special nodes */
517 #define OBD_CONNECT_LCL_CLIENT 0x00010000ULL /* local 1.8 client */
518 #define OBD_CONNECT_RMT_CLIENT 0x00020000ULL /* Remote 1.8 client */
519 #define OBD_CONNECT_BRW_SIZE 0x00040000ULL /* Max bytes per rpc */
520 #define OBD_CONNECT_QUOTA64 0x00080000ULL /* 64bit qunit_data.qd_count b=10707*/
521 #define OBD_CONNECT_MDS_CAPA 0x00100000ULL /* MDS capability */
522 #define OBD_CONNECT_OSS_CAPA 0x00200000ULL /* OSS capability */
523 #define OBD_CONNECT_CANCELSET 0x00400000ULL /* Early batched cancels. */
524 #define OBD_CONNECT_SOM 0x00800000ULL /* SOM feature */
525 #define OBD_CONNECT_AT 0x01000000ULL /* client uses adaptive timeouts */
526 #define OBD_CONNECT_LRU_RESIZE 0x02000000ULL /* Lru resize feature. */
527 #define OBD_CONNECT_MDS_MDS 0x04000000ULL /* MDS-MDS connection*/
528 #define OBD_CONNECT_REAL 0x08000000ULL /* real connection */
529 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
530 * and lustre/utils/wirecheck.c */
532 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
533 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
534 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
535 OBD_CONNECT_NODEVOH |/* OBD_CONNECT_ATTRFID |*/\
536 OBD_CONNECT_LCL_CLIENT | \
537 OBD_CONNECT_RMT_CLIENT | \
538 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | \
539 OBD_CONNECT_MDS_MDS | OBD_CONNECT_CANCELSET)
540 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
541 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
542 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
543 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
544 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET)
545 #define ECHO_CONNECT_SUPPORTED (0)
546 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION)
548 #define MAX_QUOTA_COUNT32 (0xffffffffULL)
550 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
551 ((patch)<<8) + (fix))
552 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
553 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
554 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
555 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
557 #define exp_connect_cancelset(exp) \
558 ((exp) ? (exp)->exp_connect_flags & OBD_CONNECT_CANCELSET : 0)
560 /* This structure is used for both request and reply.
562 * If we eventually have separate connect data for different types, which we
563 * almost certainly will, then perhaps we stick a union in here. */
564 struct obd_connect_data {
565 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
566 __u32 ocd_version; /* lustre release version number */
567 __u32 ocd_grant; /* initial cache grant amount (bytes) */
568 __u32 ocd_index; /* LOV index to connect to */
569 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
570 __u64 ocd_ibits_known; /* inode bits this client understands */
571 __u32 ocd_nllu; /* non-local-lustre-user */
572 __u32 ocd_nllg; /* non-local-lustre-group */
573 __u64 ocd_transno; /* first transno from client to be replayed */
574 __u32 ocd_group; /* MDS group on OST */
575 __u32 padding1; /* also fix lustre_swab_connect */
576 __u64 padding2; /* also fix lustre_swab_connect */
577 __u64 padding3; /* also fix lustre_swab_connect */
580 extern void lustre_swab_connect(struct obd_connect_data *ocd);
583 * OST requests: OBDO & OBD request records
588 OST_REPLY = 0, /* reply ? */
608 #define OST_FIRST_OPC OST_REPLY
610 typedef uint64_t obd_id;
611 typedef uint64_t obd_gr;
612 typedef uint64_t obd_time;
613 typedef uint64_t obd_size;
614 typedef uint64_t obd_off;
615 typedef uint64_t obd_blocks;
616 typedef uint32_t obd_blksize;
617 typedef uint32_t obd_mode;
618 typedef uint32_t obd_uid;
619 typedef uint32_t obd_gid;
620 typedef uint32_t obd_flag;
621 typedef uint64_t obd_valid;
622 typedef uint32_t obd_count;
624 #define OBD_FL_INLINEDATA (0x00000001)
625 #define OBD_FL_OBDMDEXISTS (0x00000002)
626 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
627 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
628 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
629 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
630 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
631 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
632 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
635 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
636 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
639 #define OBD_FL_TRUNCLOCK (0x00000800)
642 * This should not be smaller than sizeof(struct lustre_handle) + sizeof(struct
643 * llog_cookie) + sizeof(struct ll_fid). Nevertheless struct ll_fid is not
644 * longer stored in o_inline, we keep this just for case.
646 #define OBD_INLINESZ 80
648 /* Note: 64-bit types are 64-bit aligned in structure */
650 obd_valid o_valid; /* hot fields in this obdo */
654 obd_size o_size; /* o_size-o_blocks == ost_lvb */
658 obd_blocks o_blocks; /* brw: cli sent cached bytes */
661 /* 32-bit fields start here: keep an even number of them via padding */
662 obd_blksize o_blksize; /* optimal IO blocksize */
663 obd_mode o_mode; /* brw: cli sent cache remain */
667 obd_count o_nlink; /* brw: checksum */
668 obd_count o_generation;
669 obd_count o_misc; /* brw: o_dropped */
670 __u32 o_easize; /* epoch in ost writes */
672 __u32 o_stripe_idx; /* holds stripe idx */
674 char o_inline[OBD_INLINESZ];
675 /* lustre_handle + llog_cookie */
678 #define o_dirty o_blocks
679 #define o_undirty o_mode
680 #define o_dropped o_misc
681 #define o_cksum o_nlink
683 extern void lustre_swab_obdo (struct obdo *o);
686 #define LOV_MAGIC_V1 0x0BD10BD0
687 #define LOV_MAGIC LOV_MAGIC_V1
688 #define LOV_MAGIC_JOIN 0x0BD20BD0
690 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
691 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
692 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
693 #define LOV_PATTERN_CMOBD 0x200
695 #define lov_ost_data lov_ost_data_v1
696 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
697 __u64 l_object_id; /* OST object ID */
698 __u64 l_object_gr; /* OST object group (creating MDS number) */
699 __u32 l_ost_gen; /* generation of this l_ost_idx */
700 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
703 #define lov_mds_md lov_mds_md_v1
704 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
705 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
706 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
707 __u64 lmm_object_id; /* LOV object ID */
708 __u64 lmm_object_gr; /* LOV object group */
709 __u32 lmm_stripe_size; /* size of stripe in bytes */
710 __u32 lmm_stripe_count; /* num stripes in use for this object */
711 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
714 extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm);
716 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
717 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
719 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
720 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
721 #define XATTR_NAME_LOV "trusted.lov"
724 #define XATTR_NAME_LUSTRE_ACL "system.lustre_acl"
726 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
727 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
728 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
729 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
730 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
731 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
732 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
733 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
734 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
735 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
736 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
737 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
738 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
739 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
740 #define OBD_MD_FLINLINE (0x00008000ULL) /* inline data */
741 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
742 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
743 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
744 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file handle */
745 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
746 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
747 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
748 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
749 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
750 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
751 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
752 /* ->mds if epoch opens or closes */
753 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
754 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
755 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
756 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
757 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
759 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
760 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
761 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
763 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
764 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
765 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
766 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
767 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
768 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
769 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
770 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
771 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
773 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
774 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
775 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
776 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
777 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
779 static inline struct lustre_handle *obdo_handle(struct obdo *oa)
781 return (struct lustre_handle *)oa->o_inline;
784 static inline struct llog_cookie *obdo_logcookie(struct obdo *oa)
786 return (struct llog_cookie *)(oa->o_inline +
787 sizeof(struct lustre_handle));
789 /* don't forget obdo_fid which is way down at the bottom so it can
790 * come after the definition of llog_cookie */
803 __u32 os_state; /* positive error code on server */
815 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
817 /* ost_body.data values for OST_BRW */
819 #define OBD_BRW_READ 0x01
820 #define OBD_BRW_WRITE 0x02
821 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
822 #define OBD_BRW_SYNC 0x08
823 #define OBD_BRW_CHECK 0x10
824 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
825 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
826 #define OBD_BRW_DROP 0x80 /* drop the page after IO */
827 #define OBD_BRW_NOQUOTA 0x100
828 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
830 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
832 #define OST_MIN_PRECREATE 32
833 #define OST_MAX_PRECREATE 20000
842 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
844 /* multiple of 8 bytes => can array */
845 struct niobuf_remote {
851 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
853 /* request structure for OST's */
859 extern void lustre_swab_ost_body (struct ost_body *b);
860 extern void lustre_swab_ost_last_id(obd_id *id);
862 /* lock value block communicated between the filter and llite */
872 extern void lustre_swab_ost_lvb(struct ost_lvb *);
881 MDS_GETATTR_NAME = 34,
892 MDS_DONE_WRITING = 45,
903 #define MDS_FIRST_OPC MDS_GETATTR
917 } mds_reint_t, mdt_reint_t;
919 extern void lustre_swab_generic_32s (__u32 *val);
921 /* the disposition of the intent outlines what was executed */
922 #define DISP_IT_EXECD 0x00000001
923 #define DISP_LOOKUP_EXECD 0x00000002
924 #define DISP_LOOKUP_NEG 0x00000004
925 #define DISP_LOOKUP_POS 0x00000008
926 #define DISP_OPEN_CREATE 0x00000010
927 #define DISP_OPEN_OPEN 0x00000020
928 #define DISP_ENQ_COMPLETE 0x00400000
929 #define DISP_ENQ_OPEN_REF 0x00800000
930 #define DISP_ENQ_CREATE_REF 0x01000000
931 #define DISP_OPEN_LOCK 0x02000000
933 /* INODE LOCK PARTS */
934 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
935 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
936 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
938 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
939 #define MDS_INODELOCK_MAXSHIFT 2
940 /* This FULL lock is useful to take on unlink sort of operations */
941 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
944 __u64 id; /* holds object id */
945 __u32 generation; /* holds object generation */
947 __u32 f_type; /* holds object type or stripe idx when passing it to
948 * OST for saving into EA. */
951 extern void lustre_swab_ll_fid (struct ll_fid *fid);
953 #define MDS_STATUS_CONN 1
954 #define MDS_STATUS_LOV 2
956 struct mds_status_req {
961 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
963 /* mdt_thread_info.mti_flags. */
965 /* The flag indicates Size-on-MDS attributes are changed. */
966 MF_SOM_CHANGE = (1 << 0),
967 /* Flags indicates an epoch opens or closes. */
968 MF_EPOCH_OPEN = (1 << 1),
969 MF_EPOCH_CLOSE = (1 << 2),
970 MF_MDC_CANCEL_FID1 = (1 << 3),
971 MF_MDC_CANCEL_FID2 = (1 << 4),
972 MF_MDC_CANCEL_FID3 = (1 << 5),
973 MF_MDC_CANCEL_FID4 = (1 << 6),
976 #define MF_SOM_LOCAL_FLAGS (MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID2 | \
977 MF_MDC_CANCEL_FID3 | MF_MDC_CANCEL_FID4)
979 #define MDS_BFLAG_UNCOMMITTED_WRITES 0x1
980 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
982 /* these should be identical to their EXT3_*_FL counterparts, and are
983 * redefined here only to avoid dragging in ext3_fs.h */
984 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
985 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
986 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
987 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
988 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
991 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
992 * and we need to decode these into local S_* flags in the inode. Otherwise
993 * we pass flags straight through (see bug 9486). */
994 static inline int ll_ext_to_inode_flags(int flags)
996 return (flags & MDS_BFLAG_EXT_FLAGS) ?
997 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
998 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
999 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
1000 #if defined(S_DIRSYNC)
1001 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1003 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1004 (flags & ~MDS_BFLAG_EXT_FLAGS);
1007 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1008 * and we pass these straight through. Otherwise we need to convert from
1009 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1010 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1012 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1013 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1014 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1015 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1016 #if defined(S_DIRSYNC)
1017 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1019 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1026 struct lustre_handle handle;
1028 __u64 size; /* Offset, in the case of MDS_READPAGE */
1032 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1040 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1042 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1047 __u32 max_cookiesize; /* also fix lustre_swab_mdt_body */
1053 struct lustre_handle handle;
1055 __u64 size; /* Offset, in the case of MDS_READPAGE */
1059 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1068 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1070 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1076 __u32 max_cookiesize; /* also fix lustre_swab_mds_body */
1077 __u32 padding_4; /* also fix lustre_swab_mds_body */
1080 extern void lustre_swab_mds_body (struct mds_body *b);
1081 extern void lustre_swab_mdt_body (struct mdt_body *b);
1084 struct lustre_handle handle;
1090 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1092 #define Q_QUOTACHECK 0x800100
1093 #define Q_INITQUOTA 0x800101 /* init slave limits */
1094 #define Q_GETOINFO 0x800102 /* get obd quota info */
1095 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1097 #define Q_TYPESET(oqc, type) \
1098 ((oqc)->qc_type == type || (oqc)->qc_type == UGQUOTA)
1100 #define Q_GETOCMD(oqc) \
1101 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1103 struct obd_quotactl {
1108 struct obd_dqinfo qc_dqinfo;
1109 struct obd_dqblk qc_dqblk;
1112 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1114 /* inode access permission for remote user, the inode info are omitted,
1115 * for client knows them. */
1116 struct mds_remote_perm {
1121 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1124 /* setxid permissions for mds_setxid_perm.mp_perm */
1125 #define LUSTRE_SETUID_PERM 0x01
1126 #define LUSTRE_SETGID_PERM 0x02
1127 #define LUSTRE_SETGRP_PERM 0x04
1129 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1131 struct mdt_remote_perm {
1136 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1139 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1141 struct mds_rec_setattr {
1148 struct ll_fid sa_fid;
1156 __u32 sa_attr_flags;
1157 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1160 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1162 struct mdt_rec_setattr {
1169 struct lu_fid sa_fid;
1178 __u32 sa_attr_flags;
1179 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1182 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1184 /* Remove this once we declare it in include/linux/fs.h (v21 kernel patch?) */
1185 #ifndef ATTR_CTIME_SET
1186 #define ATTR_CTIME_SET 0x2000
1190 #define FMODE_READ 00000001
1191 #define FMODE_WRITE 00000002
1194 #define FMODE_EPOCH 01000000
1195 #define FMODE_EPOCHLCK 02000000
1196 #define FMODE_SOM 04000000
1197 #define FMODE_CLOSED 0
1199 #define MDS_OPEN_CREATED 00000010
1200 #define MDS_OPEN_CROSS 00000020
1202 #define MDS_FMODE_EXEC 00000004
1203 #define MDS_OPEN_CREAT 00000100
1204 #define MDS_OPEN_EXCL 00000200
1205 #define MDS_OPEN_TRUNC 00001000
1206 #define MDS_OPEN_APPEND 00002000
1207 #define MDS_OPEN_SYNC 00010000
1208 #define MDS_OPEN_DIRECTORY 00200000
1210 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1211 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1212 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1213 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1214 * with default ACL */
1215 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1216 * actually, this is for create, not
1217 * conflict with other open flags */
1218 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1219 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1220 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1222 /* permission for create non-directory file */
1223 #define MAY_CREATE (1 << 7)
1224 /* permission for create directory file */
1225 #define MAY_LINK (1 << 8)
1226 /* permission for delete from the directory */
1227 #define MAY_UNLINK (1 << 9)
1228 /* source's permission for rename */
1229 #define MAY_RENAME_SRC (1 << 10)
1230 /* target's permission for rename */
1231 #define MAY_RENAME_TAR (1 << 11)
1232 /* part (parent's) VTX permission check */
1233 #define MAY_VTX_PART (1 << 12)
1234 /* full VTX permission check */
1235 #define MAY_VTX_FULL (1 << 13)
1238 MDS_CHECK_SPLIT = 1 << 0,
1239 MDS_CROSS_REF = 1 << 1,
1240 MDS_VTX_BYPASS = 1 << 2,
1241 MDS_PERM_BYPASS = 1 << 3
1244 struct mds_rec_join {
1245 struct ll_fid jr_fid;
1249 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1251 struct mdt_rec_join {
1252 struct lu_fid jr_fid;
1256 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1258 struct mds_rec_create {
1263 __u32 cr_flags; /* for use with open */
1265 struct ll_fid cr_fid;
1266 struct ll_fid cr_replayfid;
1270 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1271 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1272 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1273 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1274 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1277 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1279 struct mdt_rec_create {
1284 __u32 cr_flags; /* for use with open */
1286 struct lustre_handle cr_old_handle; /* u64 handle in case of open replay */
1287 struct lu_fid cr_fid1;
1288 struct lu_fid cr_fid2;
1295 __u32 cr_padding_1; /* pad for 64 bits*/
1298 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1300 struct mds_rec_link {
1307 struct ll_fid lk_fid1;
1308 struct ll_fid lk_fid2;
1310 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1311 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1312 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1313 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1316 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1318 struct mdt_rec_link {
1325 struct lu_fid lk_fid1;
1326 struct lu_fid lk_fid2;
1329 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1330 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1331 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1334 extern void lustre_swab_mdt_rec_link (struct mdt_rec_link *lk);
1336 struct mds_rec_unlink {
1343 struct ll_fid ul_fid1;
1344 struct ll_fid ul_fid2;
1346 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1347 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1348 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1349 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1352 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1354 struct mdt_rec_unlink {
1361 struct lu_fid ul_fid1;
1362 struct lu_fid ul_fid2;
1365 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1366 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1367 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1370 extern void lustre_swab_mdt_rec_unlink (struct mdt_rec_unlink *ul);
1372 struct mds_rec_rename {
1379 struct ll_fid rn_fid1;
1380 struct ll_fid rn_fid2;
1382 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1383 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1384 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1385 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1388 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1390 struct mdt_rec_rename {
1397 struct lu_fid rn_fid1;
1398 struct lu_fid rn_fid2;
1400 __u32 rn_mode; /* cross-ref rename has mode */
1401 __u32 rn_bias; /* some operation flags */
1402 __u32 rn_padding_3; /* also fix lustre_swab_mdt_rec_rename */
1403 __u32 rn_padding_4; /* also fix lustre_swab_mdt_rec_rename */
1406 extern void lustre_swab_mdt_rec_rename (struct mdt_rec_rename *rn);
1408 /* begin adding MDT by huanghua@clusterfs.com */
1410 __u32 ld_tgt_count; /* how many MDS's */
1411 __u32 ld_active_tgt_count; /* how many active */
1412 struct obd_uuid ld_uuid;
1415 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1416 /* end adding MDT by huanghua@clusterfs.com */
1423 extern void lustre_swab_md_fld (struct md_fld *mf);
1428 FLD_FIRST_OPC = FLD_QUERY
1434 SEQ_FIRST_OPC = SEQ_QUERY
1438 SEQ_ALLOC_SUPER = 0,
1443 * LOV data structures
1446 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1447 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1449 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1450 /* The size of the buffer the lov/mdc reserves for the
1451 * array of UUIDs returned by the MDS. With the current
1452 * protocol, this will limit the max number of OSTs per LOV */
1454 #define LOV_DESC_MAGIC 0xB0CCDE5C
1456 /* LOV settings descriptor (should only contain static info) */
1458 __u32 ld_tgt_count; /* how many OBD's */
1459 __u32 ld_active_tgt_count; /* how many active */
1460 __u32 ld_default_stripe_count; /* how many objects are used */
1461 __u32 ld_pattern; /* default PATTERN_RAID0 */
1462 __u64 ld_default_stripe_size; /* in bytes */
1463 __u64 ld_default_stripe_offset; /* in bytes */
1464 __u32 ld_padding_0; /* unused */
1465 __u32 ld_qos_maxage; /* in second */
1466 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1467 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1468 struct obd_uuid ld_uuid;
1471 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1473 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1478 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1483 LDLM_BL_CALLBACK = 104,
1484 LDLM_CP_CALLBACK = 105,
1485 LDLM_GL_CALLBACK = 106,
1488 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1490 #define RES_NAME_SIZE 4
1491 struct ldlm_res_id {
1492 __u64 name[RES_NAME_SIZE];
1495 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1518 #define LDLM_MIN_TYPE LDLM_PLAIN
1520 struct ldlm_extent {
1526 struct ldlm_inodebits {
1533 __u64 blocking_export; /* not actually used over the wire */
1534 __u32 blocking_pid; /* not actually used over the wire */
1538 /* it's important that the fields of the ldlm_extent structure match
1539 * the first fields of the ldlm_flock structure because there is only
1540 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1541 * this ever changes we will need to swab the union differently based
1542 * on the resource type. */
1545 struct ldlm_extent l_extent;
1546 struct ldlm_flock l_flock;
1547 struct ldlm_inodebits l_inodebits;
1548 } ldlm_policy_data_t;
1550 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1552 struct ldlm_intent {
1556 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1558 struct ldlm_resource_desc {
1559 ldlm_type_t lr_type;
1560 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1561 struct ldlm_res_id lr_name;
1564 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1566 struct ldlm_lock_desc {
1567 struct ldlm_resource_desc l_resource;
1568 ldlm_mode_t l_req_mode;
1569 ldlm_mode_t l_granted_mode;
1570 ldlm_policy_data_t l_policy_data;
1573 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1575 #define LDLM_LOCKREQ_HANDLES 2
1576 #define LDLM_ENQUEUE_CANCEL_OFF 1
1578 struct ldlm_request {
1581 struct ldlm_lock_desc lock_desc;
1582 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
1585 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1587 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
1588 * Otherwise, 2 are available. */
1589 #define ldlm_request_bufsize(count,type) \
1591 int _avail = LDLM_LOCKREQ_HANDLES; \
1592 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
1593 sizeof(struct ldlm_request) + \
1594 (count > _avail ? count - _avail : 0) * \
1595 sizeof(struct lustre_handle); \
1600 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1601 struct ldlm_lock_desc lock_desc;
1602 struct lustre_handle lock_handle;
1603 __u64 lock_policy_res1;
1604 __u64 lock_policy_res2;
1607 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1610 * Opcodes for mountconf (mgs and mgc)
1615 MGS_EXCEPTION, /* node died, etc. */
1616 MGS_TARGET_REG, /* whenever target starts up */
1620 #define MGS_FIRST_OPC MGS_CONNECT
1622 /* We pass this info to the MGS so it can write config logs */
1623 #define MTI_NAME_MAXLEN 64
1624 #define MTI_PARAM_MAXLEN 4096
1625 #define MTI_NIDS_MAX 32
1626 struct mgs_target_info {
1627 __u32 mti_lustre_ver;
1628 __u32 mti_stripe_index;
1629 __u32 mti_config_ver;
1631 __u32 mti_nid_count;
1632 __u32 padding; /* 64 bit align */
1633 char mti_fsname[MTI_NAME_MAXLEN];
1634 char mti_svname[MTI_NAME_MAXLEN];
1635 char mti_uuid[sizeof(struct obd_uuid)];
1636 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
1637 char mti_params[MTI_PARAM_MAXLEN];
1640 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
1642 /* Config marker flags (in config log) */
1643 #define CM_START 0x01
1645 #define CM_SKIP 0x04
1646 #define CM_UPGRADE146 0x08
1647 #define CM_EXCLUDE 0x10
1648 #define CM_START_SKIP (CM_START | CM_SKIP)
1651 __u32 cm_step; /* aka config version */
1653 __u32 cm_vers; /* lustre release version number */
1654 __u32 padding; /* 64 bit align */
1655 time_t cm_createtime; /*when this record was first created */
1656 time_t cm_canceltime; /*when this record is no longer valid*/
1657 char cm_tgtname[MTI_NAME_MAXLEN];
1658 char cm_comment[MTI_NAME_MAXLEN];
1662 * Opcodes for multiple servers.
1671 #define OBD_FIRST_OPC OBD_PING
1673 /* catalog of log objects */
1675 /* Identifier for a single log object */
1680 } __attribute__((packed));
1682 /* Records written to the CATALOGS list */
1683 #define CATLIST "CATALOGS"
1685 struct llog_logid lci_logid;
1689 } __attribute__((packed));
1691 /*join file lov mds md*/
1692 struct lov_mds_md_join {
1693 struct lov_mds_md lmmj_md;
1694 /*join private info*/
1695 struct llog_logid lmmj_array_id; /*array object id*/
1696 __u32 lmmj_extent_count; /*array extent count*/
1699 /* Log data record types - there is no specific reason that these need to
1700 * be related to the RPC opcodes, but no reason not to (may be handy later?)
1702 #define LLOG_OP_MAGIC 0x10600000
1703 #define LLOG_OP_MASK 0xfff00000
1706 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
1707 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
1708 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
1709 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
1710 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
1711 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
1712 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
1713 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
1714 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
1715 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
1716 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
1720 * for now, continue to support old pad records which have 0 for their
1721 * type but still need to be swabbed for their length
1723 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
1724 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
1725 __swab32(LLOG_OP_MAGIC) || \
1726 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
1728 /* Log record header - stored in little endian order.
1729 * Each record must start with this struct, end with a llog_rec_tail,
1730 * and be a multiple of 256 bits in size.
1732 struct llog_rec_hdr {
1739 struct llog_rec_tail {
1744 struct llog_logid_rec {
1745 struct llog_rec_hdr lid_hdr;
1746 struct llog_logid lid_id;
1752 struct llog_rec_tail lid_tail;
1753 } __attribute__((packed));
1755 /* MDS extent description
1756 * It is for joined file extent info, each extent info for joined file
1757 * just like (start, end, lmm).
1759 struct mds_extent_desc {
1760 __u64 med_start; /* extent start */
1761 __u64 med_len; /* extent length */
1762 struct lov_mds_md med_lmm; /* extent's lmm */
1764 /*Joined file array extent log record*/
1765 struct llog_array_rec {
1766 struct llog_rec_hdr lmr_hdr;
1767 struct mds_extent_desc lmr_med;
1768 struct llog_rec_tail lmr_tail;
1771 struct llog_create_rec {
1772 struct llog_rec_hdr lcr_hdr;
1773 struct ll_fid lcr_fid;
1777 struct llog_rec_tail lcr_tail;
1778 } __attribute__((packed));
1780 struct llog_orphan_rec {
1781 struct llog_rec_hdr lor_hdr;
1785 struct llog_rec_tail lor_tail;
1786 } __attribute__((packed));
1788 struct llog_unlink_rec {
1789 struct llog_rec_hdr lur_hdr;
1793 struct llog_rec_tail lur_tail;
1794 } __attribute__((packed));
1796 struct llog_setattr_rec {
1797 struct llog_rec_hdr lsr_hdr;
1803 struct llog_rec_tail lsr_tail;
1804 } __attribute__((packed));
1806 struct llog_size_change_rec {
1807 struct llog_rec_hdr lsc_hdr;
1808 struct ll_fid lsc_fid;
1811 struct llog_rec_tail lsc_tail;
1812 } __attribute__((packed));
1817 } __attribute__((packed));
1819 struct llog_gen_rec {
1820 struct llog_rec_hdr lgr_hdr;
1821 struct llog_gen lgr_gen;
1822 struct llog_rec_tail lgr_tail;
1824 /* On-disk header structure of each log object, stored in little endian order */
1825 #define LLOG_CHUNK_SIZE 8192
1826 #define LLOG_HEADER_SIZE (96)
1827 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
1829 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
1831 /* flags for the logs */
1832 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
1833 #define LLOG_F_IS_CAT 0x2
1834 #define LLOG_F_IS_PLAIN 0x4
1836 struct llog_log_hdr {
1837 struct llog_rec_hdr llh_hdr;
1838 __u64 llh_timestamp;
1840 __u32 llh_bitmap_offset;
1844 /* for a catalog the first plain slot is next to it */
1845 struct obd_uuid llh_tgtuuid;
1846 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
1847 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
1848 struct llog_rec_tail llh_tail;
1849 } __attribute__((packed));
1851 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
1852 llh->llh_bitmap_offset - \
1853 sizeof(llh->llh_tail)) * 8)
1855 /* log cookies are used to reference a specific log file and a record therein */
1856 struct llog_cookie {
1857 struct llog_logid lgc_lgl;
1861 } __attribute__((packed));
1864 enum llogd_rpc_ops {
1865 LLOG_ORIGIN_HANDLE_CREATE = 501,
1866 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
1867 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
1868 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
1869 LLOG_ORIGIN_HANDLE_CLOSE = 505,
1870 LLOG_ORIGIN_CONNECT = 506,
1871 LLOG_CATINFO = 507, /* for lfs catinfo */
1872 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
1873 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
1875 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
1879 struct llog_logid lgd_logid;
1881 __u32 lgd_llh_flags;
1883 __u32 lgd_saved_index;
1885 __u64 lgd_cur_offset;
1886 } __attribute__((packed));
1888 struct llogd_conn_body {
1889 struct llog_gen lgdc_gen;
1890 struct llog_logid lgdc_logid;
1891 __u32 lgdc_ctxt_idx;
1892 } __attribute__((packed));
1894 struct lov_user_ost_data_join { /* per-stripe data structure */
1895 __u64 l_extent_start; /* extent start*/
1896 __u64 l_extent_end; /* extent end*/
1897 __u64 l_object_id; /* OST object ID */
1898 __u64 l_object_gr; /* OST object group (creating MDS number) */
1899 __u32 l_ost_gen; /* generation of this OST index */
1900 __u32 l_ost_idx; /* OST index in LOV */
1901 } __attribute__((packed));
1903 struct lov_user_md_join { /* LOV EA user data (host-endian) */
1904 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
1905 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1906 __u64 lmm_object_id; /* LOV object ID */
1907 __u64 lmm_object_gr; /* LOV object group */
1908 __u32 lmm_stripe_size; /* size of stripe in bytes */
1909 __u32 lmm_stripe_count; /* num stripes in use for this object */
1910 __u32 lmm_extent_count; /* extent count of lmm*/
1911 __u64 lmm_tree_id; /* mds tree object id */
1912 __u64 lmm_tree_gen; /* mds tree object gen */
1913 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
1914 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
1915 } __attribute__((packed));
1917 extern void lustre_swab_lov_user_md(struct lov_user_md *lum);
1918 extern void lustre_swab_lov_user_md_objects(struct lov_user_md *lum);
1919 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
1922 extern void lustre_swab_llogd_body (struct llogd_body *d);
1923 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
1924 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
1925 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
1926 struct llog_rec_tail *tail);
1929 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
1931 /* quota. fixed by tianzy for bug10707 */
1932 #define QUOTA_IS_GRP 0X1UL /* 0 is user, 1 is group. Used by qd_flags*/
1933 #define QUOTA_IS_BLOCK 0x2UL /* 0 is inode, 1 is block. Used by qd_flags*/
1936 __u32 qd_id; /* ID appiles to (uid, gid) */
1937 __u32 qd_flags; /* Quota type (USRQUOTA, GRPQUOTA) occupy one bit;
1938 * Block quota or file quota occupy one bit */
1939 __u64 qd_count; /* acquire/release count (bytes for block quota) */
1942 struct qunit_data_old {
1943 __u32 qd_id; /* ID appiles to (uid, gid) */
1944 __u32 qd_type; /* Quota type (USRQUOTA, GRPQUOTA) */
1945 __u32 qd_count; /* acquire/release count (bytes for block quota) */
1946 __u32 qd_isblk; /* Block quota or file quota */
1949 extern void lustre_swab_qdata(struct qunit_data *d);
1950 extern void lustre_swab_qdata_old(struct qunit_data_old *d);
1951 extern struct qunit_data *lustre_quota_old_to_new(struct qunit_data_old *d);
1952 extern struct qunit_data_old *lustre_quota_new_to_old(struct qunit_data *d);
1959 #define JOIN_FILE_ALIGN 4096
1961 /* security opcodes */
1964 SEC_CTX_INIT_CONT = 802,
1967 SEC_FIRST_OPC = SEC_CTX_INIT
1971 * capa related definitions
1973 #define CAPA_HMAC_MAX_LEN 64
1974 #define CAPA_HMAC_KEY_MAX_LEN 56
1976 /* NB take care when changing the sequence of elements this struct,
1977 * because the offset info is used in find_capa() */
1978 struct lustre_capa {
1979 struct lu_fid lc_fid; /* fid */
1980 __u64 lc_opc; /* operations allowed */
1981 __u32 lc_uid; /* uid, it is obsolete, but maybe used in
1982 * future, reserve it for 64-bits aligned.*/
1983 __u32 lc_flags; /* HMAC algorithm & flags */
1984 __u32 lc_keyid; /* key used for the capability */
1985 __u32 lc_timeout; /* capa timeout value (sec) */
1986 __u64 lc_expiry; /* expiry time (sec) */
1987 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /* HMAC */
1988 } __attribute__((packed));
1990 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
1992 /* lustre_capa.lc_opc */
1994 CAPA_OPC_BODY_WRITE = 1<<0, /* write object data */
1995 CAPA_OPC_BODY_READ = 1<<1, /* read object data */
1996 CAPA_OPC_INDEX_LOOKUP = 1<<2, /* lookup object fid */
1997 CAPA_OPC_INDEX_INSERT = 1<<3, /* insert object fid */
1998 CAPA_OPC_INDEX_DELETE = 1<<4, /* delete object fid */
1999 CAPA_OPC_OSS_WRITE = 1<<5, /* write oss object data */
2000 CAPA_OPC_OSS_READ = 1<<6, /* read oss object data */
2001 CAPA_OPC_OSS_TRUNC = 1<<7, /* truncate oss object */
2002 CAPA_OPC_META_WRITE = 1<<8, /* write object meta data */
2003 CAPA_OPC_META_READ = 1<<9, /* read object meta data */
2007 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2008 #define CAPA_OPC_MDS_ONLY \
2009 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2010 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2011 #define CAPA_OPC_OSS_ONLY \
2012 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC)
2013 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2014 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2016 /* MDS capability covers object capability for operations of body r/w
2017 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
2018 * while OSS capability only covers object capability for operations of
2019 * oss data(file content) r/w/truncate.
2021 static inline int capa_for_mds(struct lustre_capa *c)
2023 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2026 static inline int capa_for_oss(struct lustre_capa *c)
2028 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2031 /* lustre_capa.lc_hmac_alg */
2033 CAPA_HMAC_ALG_SHA1 = 1, /* sha1 algorithm */
2037 #define CAPA_FL_MASK 0x00ffffff
2038 #define CAPA_HMAC_ALG_MASK 0xff000000
2040 struct lustre_capa_key {
2041 __u64 lk_mdsid; /* mds# */
2042 __u32 lk_keyid; /* key# */
2044 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /* key */
2045 } __attribute__((packed));
2047 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2049 /* quota check function */
2050 #define QUOTA_RET_OK 0 /* return successfully */
2051 #define QUOTA_RET_NOQUOTA 1 /* not support quota */
2052 #define QUOTA_RET_NOLIMIT 2 /* quota limit isn't set */
2053 #define QUOTA_RET_ACQUOTA 3 /* need to acquire extra quota */