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.
8 * All structs passing over the wire should be declared here (lov_mds_md
9 * being the lone exception). Structs must be properly aligned to put
10 * 64-bit values on an 8-byte boundary. Any structs being added here
11 * must also be added to utils/wirecheck.c and "make newwiretest" run
12 * to regenerate the utils/wiretest.c sources. This allows us to verify
13 * that wire structs have the proper alignment/size on all architectures.
15 * We assume all nodes are either little-endian or big-endian, and we
16 * always send messages in the sender's native format. The receiver
17 * detects the message format by checking the 'magic' field of the message
18 * (see lustre_msg_swabbed() below).
20 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
21 * implemented either here, inline (trivial implementations) or in
22 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
23 * endian, in-place in the message buffer.
25 * A swabber takes a single pointer argument. The caller must already have
26 * verified that the length of the message buffer >= sizeof (type).
28 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
29 * may be defined that swabs just the variable part, after the caller has
30 * verified that the message buffer is large enough.
33 #ifndef _LUSTRE_IDL_H_
34 #define _LUSTRE_IDL_H_
36 #include <libcfs/kp30.h>
38 #if defined(__linux__)
39 #include <linux/lustre_types.h>
40 #elif defined(__APPLE__)
41 #include <darwin/lustre_types.h>
42 #elif defined(__WINNT__)
43 #include <winnt/lustre_types.h>
45 #error Unsupported operating system.
48 #include <lnet/types.h> /* for lnet_nid_t */
53 #include <libcfs/kp30.h>
55 /* Defn's shared with user-space. */
56 #include <lustre/lustre_user.h>
59 * this file contains all data structures used in Lustre interfaces:
60 * - obdo and obd_request records
61 * - mds_request records
69 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
70 * FOO_REPLY_PORTAL is for incoming replies on the FOO
71 * FOO_BULK_PORTAL is for incoming bulk on the FOO
74 #define CONNMGR_REQUEST_PORTAL 1
75 #define CONNMGR_REPLY_PORTAL 2
76 //#define OSC_REQUEST_PORTAL 3
77 #define OSC_REPLY_PORTAL 4
78 //#define OSC_BULK_PORTAL 5
79 #define OST_IO_PORTAL 6
80 #define OST_CREATE_PORTAL 7
81 #define OST_BULK_PORTAL 8
82 //#define MDC_REQUEST_PORTAL 9
83 #define MDC_REPLY_PORTAL 10
84 //#define MDC_BULK_PORTAL 11
85 #define MDS_REQUEST_PORTAL 12
86 //#define MDS_REPLY_PORTAL 13
87 #define MDS_BULK_PORTAL 14
88 #define LDLM_CB_REQUEST_PORTAL 15
89 #define LDLM_CB_REPLY_PORTAL 16
90 #define LDLM_CANCEL_REQUEST_PORTAL 17
91 #define LDLM_CANCEL_REPLY_PORTAL 18
92 //#define PTLBD_REQUEST_PORTAL 19
93 //#define PTLBD_REPLY_PORTAL 20
94 //#define PTLBD_BULK_PORTAL 21
95 #define MDS_SETATTR_PORTAL 22
96 #define MDS_READPAGE_PORTAL 23
98 #define MGC_REPLY_PORTAL 25
99 #define MGS_REQUEST_PORTAL 26
100 #define MGS_REPLY_PORTAL 27
101 #define OST_REQUEST_PORTAL 28
102 #define FLD_REQUEST_PORTAL 29
103 #define SEQ_METADATA_PORTAL 30
104 #define SEQ_DATA_PORTAL 31
105 #define SEQ_CONTROLLER_PORTAL 32
110 #define SVC_RUNNING 8
111 #define SVC_STOPPING 16
112 #define SVC_STOPPED 32
115 #define PTL_RPC_MSG_REQUEST 4711
116 #define PTL_RPC_MSG_ERR 4712
117 #define PTL_RPC_MSG_REPLY 4713
119 /* DON'T use swabbed values of MAGIC as magic! */
120 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
121 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD2
123 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
124 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD20BD00B
126 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
128 #define PTLRPC_MSG_VERSION 0x00000003
129 #define LUSTRE_VERSION_MASK 0xffff0000
130 #define LUSTRE_OBD_VERSION 0x00010000
131 #define LUSTRE_MDS_VERSION 0x00020000
132 #define LUSTRE_OST_VERSION 0x00030000
133 #define LUSTRE_DLM_VERSION 0x00040000
134 #define LUSTRE_LOG_VERSION 0x00050000
135 #define LUSTRE_MGS_VERSION 0x00060000
137 typedef __u64 mdsno_t;
138 typedef __u64 seqno_t;
145 static inline __u64 range_space(struct lu_range *r)
147 return r->lr_end - r->lr_start;
150 static inline void range_zero(struct lu_range *r)
152 r->lr_start = r->lr_end = 0;
155 static inline int range_within(struct lu_range *r,
158 return s >= r->lr_start && s < r->lr_end;
161 static inline void range_alloc(struct lu_range *r,
165 r->lr_start = s->lr_start;
166 r->lr_end = s->lr_start + w;
170 static inline int range_is_sane(struct lu_range *r)
172 return (r->lr_end >= r->lr_start);
175 static inline int range_is_zero(struct lu_range *r)
177 return (r->lr_start == 0 && r->lr_end == 0);
180 static inline int range_is_exhausted(struct lu_range *r)
182 return range_space(r) == 0;
185 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x]"
187 #define PRANGE(range) \
192 __u64 f_seq; /* holds fid sequence. Lustre should support 2 ^ 64
193 * objects, thus even if one sequence has one object we
194 * reach this value. */
195 __u32 f_oid; /* fid number within its sequence. */
196 __u32 f_ver; /* holds fid version. */
203 LUSTRE_ROOT_FID_SEQ = 1ULL, /* XXX: should go into mkfs. */
205 /* initial fid id value */
206 LUSTRE_FID_INIT_OID = 1UL
209 /* get object sequence */
210 static inline __u64 fid_seq(const struct lu_fid *fid)
216 static inline __u32 fid_oid(const struct lu_fid *fid)
221 /* get object version */
222 static inline __u32 fid_ver(const struct lu_fid *fid)
227 static inline int fid_seq_is_sane(__u64 seq)
232 static inline void fid_zero(struct lu_fid *fid)
234 memset(fid, 0, sizeof(*fid));
237 static inline int fid_is_igif(const struct lu_fid *fid)
239 return fid_seq(fid) == LUSTRE_ROOT_FID_SEQ;
242 static inline int fid_is_sane(const struct lu_fid *fid)
246 ((fid_seq_is_sane(fid_seq(fid)) && fid_oid(fid) != 0
247 && fid_ver(fid) == 0) ||
251 static inline int fid_is_zero(const struct lu_fid *fid)
253 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
256 #define DFID "[0x%16.16"LPF64"x/0x%8.8x:0x%8.8x]"
263 extern void lustre_swab_lu_fid(struct lu_fid *fid);
264 extern void lustre_swab_lu_range(struct lu_range *range);
266 static inline int lu_fid_eq(const struct lu_fid *f0,
267 const struct lu_fid *f1)
269 /* Check that there is no alignment padding. */
270 CLASSERT(sizeof *f0 ==
271 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
272 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
273 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
274 return memcmp(f0, f1, sizeof *f0) == 0;
278 * Layout of readdir pages, as transmitted on wire.
281 struct lu_fid lde_fid;
289 __u32 ldp_hash_start;
293 struct lu_dirent ldp_entries[0];
296 enum lu_dirpage_flags {
300 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
302 if (le16_to_cpu(dp->ldp_flags) & LDF_EMPTY)
305 return dp->ldp_entries;
308 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
310 struct lu_dirent *next;
312 if (le16_to_cpu(ent->lde_reclen) != 0)
313 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
320 static inline int lu_dirent_size(struct lu_dirent *ent)
322 if (le16_to_cpu(ent->lde_reclen) == 0) {
323 return (sizeof(*ent) +
324 le16_to_cpu(ent->lde_namelen) + 3) & ~3;
326 return le16_to_cpu(ent->lde_reclen);
329 #define DIR_END_OFF 0xfffffffeUL
331 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
332 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
333 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
335 #define MAX_HASH_SIZE 0x7fffffffUL
336 #define MAX_HASH_HIGHEST_BIT 0x10000000
338 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
339 struct lmv_stripe_md {
343 struct lu_fid mea_ids[0];
346 struct lustre_handle {
349 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
351 static inline int lustre_handle_is_used(struct lustre_handle *lh)
353 return lh->cookie != 0ull;
356 static inline int lustre_handle_equal(struct lustre_handle *lh1,
357 struct lustre_handle *lh2)
359 return lh1->cookie == lh2->cookie;
362 static inline void lustre_handle_copy(struct lustre_handle *tgt,
363 struct lustre_handle *src)
365 tgt->cookie = src->cookie;
368 /* we depend on this structure to be 8-byte aligned */
369 /* this type is only endian-adjusted in lustre_unpack_msg() */
370 struct lustre_msg_v1 {
371 struct lustre_handle lm_handle;
377 __u64 lm_last_committed;
386 #define lustre_msg lustre_msg_v2
387 /* we depend on this structure to be 8-byte aligned */
388 /* this type is only endian-adjusted in lustre_unpack_msg() */
389 struct lustre_msg_v2 {
397 /* without gss, ptlrpc_body is put at the first buffer. */
399 struct lustre_handle pb_handle;
405 __u64 pb_last_committed;
410 __u32 pb_paddings[3];
413 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
415 /* message body offset for lustre_msg_v2 */
416 /* ptlrpc body offset in all request/reply messages */
417 #define MSG_PTLRPC_BODY_OFF 0
419 /* normal request/reply message record offset */
420 #define REQ_REC_OFF 1
421 #define REPLY_REC_OFF 1
423 /* ldlm request message body offset */
424 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
425 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
427 /* ldlm intent lock message body offset */
428 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
429 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
431 /* ldlm reply message body offset */
432 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
433 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
435 /* Flags that are operation-specific go in the top 16 bits. */
436 #define MSG_OP_FLAG_MASK 0xffff0000
437 #define MSG_OP_FLAG_SHIFT 16
439 /* Flags that apply to all requests are in the bottom 16 bits */
440 #define MSG_GEN_FLAG_MASK 0x0000ffff
441 #define MSG_LAST_REPLAY 1
444 #define MSG_REQ_REPLAY_DONE 8
445 #define MSG_LOCK_REPLAY_DONE 16
448 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
451 #define MSG_CONNECT_RECOVERING 0x1
452 #define MSG_CONNECT_RECONNECT 0x2
453 #define MSG_CONNECT_REPLAYABLE 0x4
454 //#define MSG_CONNECT_PEER 0x8
455 #define MSG_CONNECT_LIBCLIENT 0x10
456 #define MSG_CONNECT_INITIAL 0x20
457 #define MSG_CONNECT_ASYNC 0x40
458 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
459 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
462 #define OBD_CONNECT_RDONLY 0x1ULL /* client allowed read-only access */
463 #define OBD_CONNECT_INDEX 0x2ULL /* connect to specific LOV idx */
464 #define OBD_CONNECT_GRANT 0x8ULL /* OSC acquires grant at connect */
465 #define OBD_CONNECT_SRVLOCK 0x10ULL /* server takes locks for client */
466 #define OBD_CONNECT_VERSION 0x20ULL /* Server supports versions in ocd */
467 #define OBD_CONNECT_REQPORTAL 0x40ULL /* Separate portal for non-IO reqs */
468 #define OBD_CONNECT_ACL 0x80ULL /* client uses access control lists */
469 #define OBD_CONNECT_XATTR 0x100ULL /* client using extended attributes*/
470 #define OBD_CONNECT_CROW 0x200ULL /* MDS+OST create objects on write */
471 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /* locks on server for punch b=9528 */
472 #define OBD_CONNECT_TRANSNO 0x800ULL /* replay sends initial transno */
473 #define OBD_CONNECT_IBITS 0x1000ULL /* support for inodebits locks */
474 #define OBD_CONNECT_JOIN 0x2000ULL /* files can be concatenated */
475 #define OBD_CONNECT_REAL 0x4000ULL
476 #define OBD_CONNECT_ATTRFID 0x8000ULL /* Server supports GetAttr By Fid */
477 #define OBD_CONNECT_NODEVOH 0x10000ULL /* No open handle for special nodes */
478 #define OBD_CONNECT_LCL_CLIENT 0x20000ULL /* local 1.8 client */
479 #define OBD_CONNECT_RMT_CLIENT 0x40000ULL /* Remote 1.8 client */
480 #define OBD_CONNECT_BRW_SIZE 0x80000ULL /* Max bytes per rpc */
481 #define OBD_CONNECT_QUOTA64 0x100000ULL /* 64bit qunit_data.qd_count b=10707*/
482 #define OBD_CONNECT_MDS_CAPA 0x200000ULL /* MDS capability */
483 #define OBD_CONNECT_OSS_CAPA 0x400000ULL /* OSS capability */
484 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
485 * and lustre/utils/wirecheck.c */
487 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
488 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
489 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
490 OBD_CONNECT_NODEVOH | OBD_CONNECT_ATTRFID | \
491 OBD_CONNECT_LCL_CLIENT | \
492 OBD_CONNECT_RMT_CLIENT | \
493 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA)
494 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
495 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
496 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
497 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
498 OBD_CONNECT_OSS_CAPA)
499 #define ECHO_CONNECT_SUPPORTED (0)
500 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION)
502 #define MAX_QUOTA_COUNT32 ((0xffffffffULL >> QUOTABLOCK_BITS) << QUOTABLOCK_BITS)
504 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
505 ((patch)<<8) + (fix))
506 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
507 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
508 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
509 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
511 /* This structure is used for both request and reply.
513 * If we eventually have separate connect data for different types, which we
514 * almost certainly will, then perhaps we stick a union in here. */
515 struct obd_connect_data {
516 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
517 __u64 ocd_transno; /* first transno from client to be replayed */
518 __u64 ocd_ibits_known; /* inode bits this client understands */
519 __u32 ocd_version; /* lustre release version number */
520 __u32 ocd_grant; /* initial cache grant amount (bytes) */
521 __u32 ocd_index; /* LOV index to connect to */
522 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
523 __u32 ocd_nllu; /* non-local-lustre-user */
524 __u32 ocd_nllg; /* non-local-lustre-group */
525 __u32 ocd_group; /* MDS group on OST */
526 __u32 padding1; /* also fix lustre_swab_connect */
527 __u64 padding2; /* also fix lustre_swab_connect */
528 __u64 padding3; /* also fix lustre_swab_connect */
531 extern void lustre_swab_connect(struct obd_connect_data *ocd);
534 * OST requests: OBDO & OBD request records
539 OST_REPLY = 0, /* reply ? */
553 /* OST_SAN_READ = 14, deprecated */
554 /* OST_SAN_WRITE = 15, deprecated */
561 #define OST_FIRST_OPC OST_REPLY
563 typedef uint64_t obd_id;
564 typedef uint64_t obd_gr;
565 typedef uint64_t obd_time;
566 typedef uint64_t obd_size;
567 typedef uint64_t obd_off;
568 typedef uint64_t obd_blocks;
569 typedef uint32_t obd_blksize;
570 typedef uint32_t obd_mode;
571 typedef uint32_t obd_uid;
572 typedef uint32_t obd_gid;
573 typedef uint32_t obd_flag;
574 typedef uint64_t obd_valid;
575 typedef uint32_t obd_count;
577 #define OBD_FL_INLINEDATA (0x00000001)
578 #define OBD_FL_OBDMDEXISTS (0x00000002)
579 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
580 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
581 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
582 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
583 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
584 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
585 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
586 #define OBD_FL_CREATE_CROW (0x00000400) /* object should be create on write */
589 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
590 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
593 #define OBD_FL_TRUNCLOCK (0x00000800)
596 * This should not be smaller than sizeof(struct lustre_handle) + sizeof(struct
597 * llog_cookie) + sizeof(struct ll_fid). Nevertheless struct ll_fid is not
598 * longer stored in o_inline, we keep this just for case.
600 #define OBD_INLINESZ 80
602 /* Note: 64-bit types are 64-bit aligned in structure */
604 obd_valid o_valid; /* hot fields in this obdo */
608 obd_size o_size; /* o_size-o_blocks == ost_lvb */
612 obd_blocks o_blocks; /* brw: cli sent cached bytes */
615 /* 32-bit fields start here: keep an even number of them via padding */
616 obd_blksize o_blksize; /* optimal IO blocksize */
617 obd_mode o_mode; /* brw: cli sent cache remain */
621 obd_count o_nlink; /* brw: checksum */
622 obd_count o_generation;
623 obd_count o_misc; /* brw: o_dropped */
624 __u32 o_easize; /* epoch in ost writes */
626 __u32 o_stripe_idx; /* holds stripe idx */
628 char o_inline[OBD_INLINESZ];
629 /* lustre_handle + llog_cookie */
632 #define o_dirty o_blocks
633 #define o_undirty o_mode
634 #define o_dropped o_misc
635 #define o_cksum o_nlink
637 extern void lustre_swab_obdo (struct obdo *o);
640 LUSTRE_OPC_MKDIR = (1 << 0),
641 LUSTRE_OPC_SYMLINK = (1 << 1),
642 LUSTRE_OPC_MKNOD = (1 << 2),
643 LUSTRE_OPC_CREATE = (1 << 3),
644 LUSTRE_OPC_ANY = (1 << 4)
648 struct lu_fid op_fid1;
649 struct lu_fid op_fid2;
650 struct lustre_handle op_handle;
655 struct lmv_stripe_md *op_mea1;
656 struct lmv_stripe_md *op_mea2;
657 __u32 op_suppgids[2];
662 /* iattr fields and blocks. */
663 struct iattr op_attr;
665 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
666 unsigned int op_attr_flags;
669 loff_t op_attr_blocks;
671 /* Size-on-MDS epoch and flags. */
676 struct obd_capa *op_capa1;
677 struct obd_capa *op_capa2;
679 /* Various operation flags. */
686 #define MDS_MODE_DONT_LOCK (1 << 30)
687 #define MDS_MODE_REPLAY (1 << 31)
689 #define LOV_MAGIC_V1 0x0BD10BD0
690 #define LOV_MAGIC LOV_MAGIC_V1
691 #define LOV_MAGIC_JOIN 0x0BD20BD0
693 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
694 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
695 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
696 #define LOV_PATTERN_CMOBD 0x200
698 #define lov_ost_data lov_ost_data_v1
699 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
700 __u64 l_object_id; /* OST object ID */
701 __u64 l_object_gr; /* OST object group (creating MDS number) */
702 __u32 l_ost_gen; /* generation of this l_ost_idx */
703 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
706 #define lov_mds_md lov_mds_md_v1
707 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
708 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
709 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
710 __u64 lmm_object_id; /* LOV object ID */
711 __u64 lmm_object_gr; /* LOV object group */
712 __u32 lmm_stripe_size; /* size of stripe in bytes */
713 __u32 lmm_stripe_count; /* num stripes in use for this object */
714 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
717 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
718 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
720 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
721 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
722 #define XATTR_NAME_LOV "trusted.lov"
725 #define XATTR_NAME_LUSTRE_ACL "system.lustre_acl"
727 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
728 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
729 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
730 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
731 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
732 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
733 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
734 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
735 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
736 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
737 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
738 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
739 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
740 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
741 #define OBD_MD_FLINLINE (0x00008000ULL) /* inline data */
742 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
743 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
744 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
745 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file handle */
746 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
747 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
748 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
749 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
750 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
751 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
752 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
753 /* ->mds if epoch opens or closes */
754 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
755 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
756 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
757 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
758 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
760 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
761 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
762 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD EA */
764 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
765 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
766 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
767 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
768 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
769 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
770 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
771 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
772 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
774 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
775 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
776 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
777 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
778 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
780 static inline struct lustre_handle *obdo_handle(struct obdo *oa)
782 return (struct lustre_handle *)oa->o_inline;
785 static inline struct llog_cookie *obdo_logcookie(struct obdo *oa)
787 return (struct llog_cookie *)(oa->o_inline +
788 sizeof(struct lustre_handle));
790 /* don't forget obdo_fid which is way down at the bottom so it can
791 * come after the definition of llog_cookie */
804 __u32 os_state; /* positive error code on server */
816 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
818 /* ost_body.data values for OST_BRW */
820 #define OBD_BRW_READ 0x01
821 #define OBD_BRW_WRITE 0x02
822 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
823 #define OBD_BRW_SYNC 0x08
824 #define OBD_BRW_CHECK 0x10
825 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
826 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
827 #define OBD_BRW_DROP 0x80 /* drop the page after IO */
828 #define OBD_BRW_NOQUOTA 0x100
829 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
831 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
833 #define OST_MIN_PRECREATE 32
834 #define OST_MAX_PRECREATE 20000
843 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
845 /* multiple of 8 bytes => can array */
846 struct niobuf_remote {
852 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
854 /* request structure for OST's */
860 extern void lustre_swab_ost_body (struct ost_body *b);
861 extern void lustre_swab_ost_last_id(obd_id *id);
862 extern void lustre_swab_generic_32s(__u32 *val);
864 /* lock value block communicated between the filter and llite */
874 extern void lustre_swab_ost_lvb(struct ost_lvb *);
883 MDS_GETATTR_NAME = 34,
894 MDS_DONE_WRITING = 45,
905 #define MDS_FIRST_OPC MDS_GETATTR
921 } mds_reint_t, mdt_reint_t;
923 /* the disposition of the intent outlines what was executed */
924 #define DISP_IT_EXECD 0x00000001
925 #define DISP_LOOKUP_EXECD 0x00000002
926 #define DISP_LOOKUP_NEG 0x00000004
927 #define DISP_LOOKUP_POS 0x00000008
928 #define DISP_OPEN_CREATE 0x00000010
929 #define DISP_OPEN_OPEN 0x00000020
930 #define DISP_ENQ_COMPLETE 0x00400000
931 #define DISP_ENQ_OPEN_REF 0x00800000
932 #define DISP_ENQ_CREATE_REF 0x01000000
934 /* INODE LOCK PARTS */
935 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
936 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
937 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
939 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
940 #define MDS_INODELOCK_MAXSHIFT 2
941 /* This FULL lock is useful to take on unlink sort of operations */
942 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
944 #define LUSTRE_CONFIG_SET 0
945 #define LUSTRE_CONFIG_GET 1
947 #define LUSTRE_CONFIG_METASEQ "metaseq"
948 #define LUSTRE_CONFIG_TRANSNO "transno"
951 __u64 id; /* holds object id */
952 __u32 generation; /* holds object generation */
954 __u32 f_type; /* holds object type or stripe idx when passing it to
955 * OST for saving into EA. */
958 extern void lustre_swab_ll_fid (struct ll_fid *fid);
960 #define MDS_STATUS_CONN 1
961 #define MDS_STATUS_LOV 2
963 struct mds_status_req {
968 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
970 /* mdt_thread_info.mti_flags. */
971 enum mdt_ioepoch_flags {
972 /* The flag indicates Size-on-MDS attributes are changed. */
973 MF_SOM_CHANGE = (1 << 0),
974 /* Flags indicates an epoch opens or closes. */
975 MF_EPOCH_OPEN = (1 << 1),
976 MF_EPOCH_CLOSE = (1 << 2),
979 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
981 /* these should be identical to their EXT3_*_FL counterparts, and are
982 * redefined here only to avoid dragging in ext3_fs.h */
983 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
984 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
985 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
986 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
987 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
990 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
991 * and we need to decode these into local S_* flags in the inode. Otherwise
992 * we pass flags straight through (see bug 9486). */
993 static inline int ll_ext_to_inode_flags(int flags)
995 return (flags & MDS_BFLAG_EXT_FLAGS) ?
996 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
997 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
998 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
999 #if defined(S_DIRSYNC)
1000 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1002 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1003 (flags & ~MDS_BFLAG_EXT_FLAGS);
1006 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1007 * and we pass these straight through. Otherwise we need to convert from
1008 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1009 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1011 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1012 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1013 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1014 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1015 #if defined(S_DIRSYNC)
1016 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1018 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1025 struct lustre_handle handle;
1027 __u64 size; /* Offset, in the case of MDS_READPAGE */
1031 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1039 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1041 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1046 __u32 max_cookiesize; /* also fix lustre_swab_mdt_body */
1052 struct lustre_handle handle;
1054 __u64 size; /* Offset, in the case of MDS_READPAGE */
1058 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1067 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1069 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1075 __u32 max_cookiesize; /* also fix lustre_swab_mds_body */
1076 __u32 padding_4; /* also fix lustre_swab_mds_body */
1079 extern void lustre_swab_mds_body (struct mds_body *b);
1080 extern void lustre_swab_mdt_body (struct mdt_body *b);
1083 struct lustre_handle handle;
1088 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1091 struct mdt_body *body;
1092 struct lov_stripe_md *lsm;
1093 struct lmv_stripe_md *mea;
1094 #ifdef CONFIG_FS_POSIX_ACL
1095 struct posix_acl *posix_acl;
1097 struct mdt_remote_perm *remote_perm;
1098 struct obd_capa *mds_capa;
1099 struct obd_capa *oss_capa;
1102 #define Q_QUOTACHECK 0x800100
1103 #define Q_INITQUOTA 0x800101 /* init slave limits */
1104 #define Q_GETOINFO 0x800102 /* get obd quota info */
1105 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1107 #define Q_TYPESET(oqc, type) \
1108 ((oqc)->qc_type == type || (oqc)->qc_type == UGQUOTA)
1110 #define Q_GETOCMD(oqc) \
1111 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1113 struct obd_quotactl {
1118 struct obd_dqinfo qc_dqinfo;
1119 struct obd_dqblk qc_dqblk;
1122 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1124 /* inode access permission for remote user, the inode info are omitted,
1125 * for client knows them. */
1126 struct mds_remote_perm {
1131 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1134 /* setxid permissions for mds_setxid_perm.mp_perm */
1135 #define LUSTRE_SETUID_PERM 0x01
1136 #define LUSTRE_SETGID_PERM 0x02
1137 #define LUSTRE_SETGRP_PERM 0x04
1139 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1141 struct mdt_remote_perm {
1146 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1149 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1151 struct mds_rec_setattr {
1158 struct ll_fid sa_fid;
1166 __u32 sa_attr_flags;
1167 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1170 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1172 struct mdt_rec_setattr {
1179 struct lu_fid sa_fid;
1188 __u32 sa_attr_flags;
1189 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1192 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1194 /* Remove this once we declare it in include/linux/fs.h (v21 kernel patch?) */
1195 #ifndef ATTR_CTIME_SET
1196 #define ATTR_CTIME_SET 0x2000
1200 #define FMODE_READ 00000001
1201 #define FMODE_WRITE 00000002
1204 #define FMODE_EPOCH 01000000
1205 #define FMODE_EPOCHLCK 02000000
1206 #define FMODE_SOM 04000000
1207 #define FMODE_CLOSED 0
1209 #define MDS_OPEN_CREATED 00000010
1211 #define MDS_FMODE_EXEC 00000004
1212 #define MDS_OPEN_CREAT 00000100
1213 #define MDS_OPEN_EXCL 00000200
1214 #define MDS_OPEN_TRUNC 00001000
1215 #define MDS_OPEN_APPEND 00002000
1216 #define MDS_OPEN_SYNC 00010000
1217 #define MDS_OPEN_DIRECTORY 00200000
1219 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1220 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1221 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1222 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1223 * with default ACL */
1224 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1225 * actually, this is for create, not
1226 * conflict with other open flags */
1227 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1228 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1229 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1231 /* permission for increase nlink of a file */
1235 MDS_CHECK_SPLIT = 1 << 0,
1236 MDS_CROSS_REF = 1 << 1
1239 struct mds_rec_join {
1240 struct ll_fid jr_fid;
1244 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1246 struct mdt_rec_join {
1247 struct lu_fid jr_fid;
1251 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1253 struct mds_rec_create {
1258 __u32 cr_flags; /* for use with open */
1260 struct ll_fid cr_fid;
1261 struct ll_fid cr_replayfid;
1265 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1266 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1267 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1268 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1269 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1272 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1274 struct mdt_rec_create {
1279 __u32 cr_flags; /* for use with open */
1281 struct lu_fid cr_fid1;
1282 struct lu_fid cr_fid2;
1288 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1289 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1292 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1294 struct mds_rec_link {
1301 struct ll_fid lk_fid1;
1302 struct ll_fid lk_fid2;
1304 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1305 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1306 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1307 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1310 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1312 struct mdt_rec_link {
1319 struct lu_fid lk_fid1;
1320 struct lu_fid lk_fid2;
1323 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1324 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1325 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1328 extern void lustre_swab_mdt_rec_link (struct mdt_rec_link *lk);
1330 struct mds_rec_unlink {
1337 struct ll_fid ul_fid1;
1338 struct ll_fid ul_fid2;
1340 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1341 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1342 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1343 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1346 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1348 struct mdt_rec_unlink {
1355 struct lu_fid ul_fid1;
1356 struct lu_fid ul_fid2;
1359 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1360 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1361 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1364 extern void lustre_swab_mdt_rec_unlink (struct mdt_rec_unlink *ul);
1366 struct mds_rec_rename {
1373 struct ll_fid rn_fid1;
1374 struct ll_fid rn_fid2;
1376 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1377 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1378 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1379 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1382 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1384 struct mdt_rec_rename {
1391 struct lu_fid rn_fid1;
1392 struct lu_fid rn_fid2;
1394 __u32 rn_mode; /* cross-ref rename has mode */
1395 __u32 rn_bias; /* some operation flags */
1396 __u32 rn_padding_3; /* also fix lustre_swab_mdt_rec_rename */
1397 __u32 rn_padding_4; /* also fix lustre_swab_mdt_rec_rename */
1400 extern void lustre_swab_mdt_rec_rename (struct mdt_rec_rename *rn);
1402 /* begin adding MDT by huanghua@clusterfs.com */
1404 __u32 ld_tgt_count; /* how many MDS's */
1405 __u32 ld_active_tgt_count; /* how many active */
1406 struct obd_uuid ld_uuid;
1409 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1410 /* end adding MDT by huanghua@clusterfs.com */
1417 extern void lustre_swab_md_fld (struct md_fld *mf);
1422 FLD_FIRST_OPC = FLD_QUERY
1428 SEQ_FIRST_OPC = SEQ_QUERY
1432 SEQ_ALLOC_SUPER = 0,
1437 * LOV data structures
1440 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1441 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1443 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1444 /* The size of the buffer the lov/mdc reserves for the
1445 * array of UUIDs returned by the MDS. With the current
1446 * protocol, this will limit the max number of OSTs per LOV */
1448 #define LOV_DESC_MAGIC 0xB0CCDE5C
1450 /* LOV settings descriptor (should only contain static info) */
1452 __u32 ld_tgt_count; /* how many OBD's */
1453 __u32 ld_active_tgt_count; /* how many active */
1454 __u32 ld_default_stripe_count; /* how many objects are used */
1455 __u32 ld_pattern; /* default PATTERN_RAID0 */
1456 __u64 ld_default_stripe_size; /* in bytes */
1457 __u64 ld_default_stripe_offset; /* in bytes */
1458 __u32 ld_padding_0; /* unused */
1459 __u32 ld_qos_maxage; /* in second */
1460 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1461 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1462 struct obd_uuid ld_uuid;
1465 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1467 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1472 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1477 LDLM_BL_CALLBACK = 104,
1478 LDLM_CP_CALLBACK = 105,
1479 LDLM_GL_CALLBACK = 106,
1482 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1484 #define RES_NAME_SIZE 4
1485 struct ldlm_res_id {
1486 __u64 name[RES_NAME_SIZE];
1489 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1512 #define LDLM_MIN_TYPE LDLM_PLAIN
1514 struct ldlm_extent {
1520 struct ldlm_inodebits {
1527 __u64 blocking_export; /* not actually used over the wire */
1528 __u32 blocking_pid; /* not actually used over the wire */
1532 /* it's important that the fields of the ldlm_extent structure match
1533 * the first fields of the ldlm_flock structure because there is only
1534 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1535 * this ever changes we will need to swab the union differently based
1536 * on the resource type. */
1539 struct ldlm_extent l_extent;
1540 struct ldlm_flock l_flock;
1541 struct ldlm_inodebits l_inodebits;
1542 } ldlm_policy_data_t;
1544 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1546 struct ldlm_intent {
1550 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1552 struct ldlm_resource_desc {
1553 ldlm_type_t lr_type;
1554 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1555 struct ldlm_res_id lr_name;
1558 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1560 struct ldlm_lock_desc {
1561 struct ldlm_resource_desc l_resource;
1562 ldlm_mode_t l_req_mode;
1563 ldlm_mode_t l_granted_mode;
1564 ldlm_policy_data_t l_policy_data;
1567 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1569 struct ldlm_request {
1571 __u32 lock_padding; /* also fix lustre_swab_ldlm_request */
1572 struct ldlm_lock_desc lock_desc;
1573 struct lustre_handle lock_handle1;
1574 struct lustre_handle lock_handle2;
1577 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1581 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1582 struct ldlm_lock_desc lock_desc;
1583 struct lustre_handle lock_handle;
1584 __u64 lock_policy_res1;
1585 __u64 lock_policy_res2;
1588 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1592 * Opcodes for mountconf (mgs and mgc)
1597 MGS_EXCEPTION, /* node died, etc. */
1598 MGS_TARGET_REG, /* whenever target starts up */
1603 /* We pass this info to the MGS so it can write config logs */
1604 #define MTI_NAME_MAXLEN 64
1605 #define MTI_NIDS_MAX 32
1606 struct mgs_target_info {
1607 __u32 mti_lustre_ver;
1608 __u32 mti_stripe_index;
1609 __u32 mti_config_ver;
1611 __u32 mti_nid_count;
1612 __u32 padding; /* 64 bit align */
1613 char mti_fsname[MTI_NAME_MAXLEN];
1614 char mti_svname[MTI_NAME_MAXLEN];
1615 char mti_uuid[sizeof(struct obd_uuid)];
1616 lnet_nid_t mti_nids[MTI_NIDS_MAX]; /* host nids */
1617 char mti_params[2048];
1620 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
1622 #define CM_START 0x01
1624 #define CM_SKIP 0x04
1625 #define CM_UPGRADE146 0x08
1626 #define CM_START_SKIP (CM_START | CM_SKIP)
1629 __u32 cm_step; /* aka config version */
1631 __u32 cm_vers; /* lustre release version number */
1632 __u32 padding; /* 64 bit align */
1633 time_t cm_createtime; /*when this record was first created */
1634 time_t cm_canceltime; /*when this record is no longer valid*/
1635 char cm_svname[MTI_NAME_MAXLEN];
1636 char cm_comment[MTI_NAME_MAXLEN];
1640 * Opcodes for multiple servers.
1649 #define OBD_FIRST_OPC OBD_PING
1651 /* catalog of log objects */
1653 /* Identifier for a single log object */
1658 } __attribute__((packed));
1660 /* Records written to the CATALOGS list */
1661 #define CATLIST "CATALOGS"
1663 struct llog_logid lci_logid;
1667 } __attribute__((packed));
1669 /*join file lov mds md*/
1670 struct lov_mds_md_join {
1671 struct lov_mds_md lmmj_md;
1672 /*join private info*/
1673 struct llog_logid lmmj_array_id; /*array object id*/
1674 __u32 lmmj_extent_count; /*array extent count*/
1677 /* Log data record types - there is no specific reason that these need to
1678 * be related to the RPC opcodes, but no reason not to (may be handy later?)
1680 #define LLOG_OP_MAGIC 0x10600000
1681 #define LLOG_OP_MASK 0xfff00000
1684 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
1685 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
1686 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
1687 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
1688 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
1689 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
1690 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
1691 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
1692 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
1693 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
1694 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
1698 * for now, continue to support old pad records which have 0 for their
1699 * type but still need to be swabbed for their length
1701 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
1702 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
1703 __swab32(LLOG_OP_MAGIC) || \
1704 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
1706 /* Log record header - stored in little endian order.
1707 * Each record must start with this struct, end with a llog_rec_tail,
1708 * and be a multiple of 256 bits in size.
1710 struct llog_rec_hdr {
1717 struct llog_rec_tail {
1722 struct llog_logid_rec {
1723 struct llog_rec_hdr lid_hdr;
1724 struct llog_logid lid_id;
1730 struct llog_rec_tail lid_tail;
1731 } __attribute__((packed));
1733 /* MDS extent description
1734 * It is for joined file extent info, each extent info for joined file
1735 * just like (start, end, lmm).
1737 struct mds_extent_desc {
1738 __u64 med_start; /* extent start */
1739 __u64 med_len; /* extent length */
1740 struct lov_mds_md med_lmm; /* extent's lmm */
1742 /*Joined file array extent log record*/
1743 struct llog_array_rec {
1744 struct llog_rec_hdr lmr_hdr;
1745 struct mds_extent_desc lmr_med;
1746 struct llog_rec_tail lmr_tail;
1749 struct llog_create_rec {
1750 struct llog_rec_hdr lcr_hdr;
1751 struct ll_fid lcr_fid;
1755 struct llog_rec_tail lcr_tail;
1756 } __attribute__((packed));
1758 struct llog_orphan_rec {
1759 struct llog_rec_hdr lor_hdr;
1763 struct llog_rec_tail lor_tail;
1764 } __attribute__((packed));
1766 struct llog_unlink_rec {
1767 struct llog_rec_hdr lur_hdr;
1771 struct llog_rec_tail lur_tail;
1772 } __attribute__((packed));
1774 struct llog_setattr_rec {
1775 struct llog_rec_hdr lsr_hdr;
1781 struct llog_rec_tail lsr_tail;
1782 } __attribute__((packed));
1784 struct llog_size_change_rec {
1785 struct llog_rec_hdr lsc_hdr;
1786 struct ll_fid lsc_fid;
1789 struct llog_rec_tail lsc_tail;
1790 } __attribute__((packed));
1795 } __attribute__((packed));
1797 struct llog_gen_rec {
1798 struct llog_rec_hdr lgr_hdr;
1799 struct llog_gen lgr_gen;
1800 struct llog_rec_tail lgr_tail;
1802 /* On-disk header structure of each log object, stored in little endian order */
1803 #define LLOG_CHUNK_SIZE 8192
1804 #define LLOG_HEADER_SIZE (96)
1805 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
1807 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
1809 /* flags for the logs */
1810 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
1811 #define LLOG_F_IS_CAT 0x2
1812 #define LLOG_F_IS_PLAIN 0x4
1814 struct llog_log_hdr {
1815 struct llog_rec_hdr llh_hdr;
1816 __u64 llh_timestamp;
1818 __u32 llh_bitmap_offset;
1822 /* for a catalog the first plain slot is next to it */
1823 struct obd_uuid llh_tgtuuid;
1824 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
1825 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
1826 struct llog_rec_tail llh_tail;
1827 } __attribute__((packed));
1829 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
1830 llh->llh_bitmap_offset - \
1831 sizeof(llh->llh_tail)) * 8)
1833 /* log cookies are used to reference a specific log file and a record therein */
1834 struct llog_cookie {
1835 struct llog_logid lgc_lgl;
1839 } __attribute__((packed));
1842 enum llogd_rpc_ops {
1843 LLOG_ORIGIN_HANDLE_CREATE = 501,
1844 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
1845 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
1846 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
1847 LLOG_ORIGIN_HANDLE_CLOSE = 505,
1848 LLOG_ORIGIN_CONNECT = 506,
1849 LLOG_CATINFO = 507, /* for lfs catinfo */
1850 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
1851 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
1856 struct llog_logid lgd_logid;
1858 __u32 lgd_llh_flags;
1860 __u32 lgd_saved_index;
1862 __u64 lgd_cur_offset;
1863 } __attribute__((packed));
1865 struct llogd_conn_body {
1866 struct llog_gen lgdc_gen;
1867 struct llog_logid lgdc_logid;
1868 __u32 lgdc_ctxt_idx;
1869 } __attribute__((packed));
1871 struct lov_user_ost_data_join { /* per-stripe data structure */
1872 __u64 l_extent_start; /* extent start*/
1873 __u64 l_extent_end; /* extent end*/
1874 __u64 l_object_id; /* OST object ID */
1875 __u64 l_object_gr; /* OST object group (creating MDS number) */
1876 __u32 l_ost_gen; /* generation of this OST index */
1877 __u32 l_ost_idx; /* OST index in LOV */
1878 } __attribute__((packed));
1880 struct lov_user_md_join { /* LOV EA user data (host-endian) */
1881 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
1882 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1883 __u64 lmm_object_id; /* LOV object ID */
1884 __u64 lmm_object_gr; /* LOV object group */
1885 __u32 lmm_stripe_size; /* size of stripe in bytes */
1886 __u32 lmm_stripe_count; /* num stripes in use for this object */
1887 __u32 lmm_extent_count; /* extent count of lmm*/
1888 __u64 lmm_tree_id; /* mds tree object id */
1889 __u64 lmm_tree_gen; /* mds tree object gen */
1890 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
1891 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
1892 } __attribute__((packed));
1894 extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm);
1895 extern void lustre_swab_lov_user_md(struct lov_user_md *lum);
1896 extern void lustre_swab_lov_user_md_objects(struct lov_user_md *lum);
1897 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
1900 extern void lustre_swab_llogd_body (struct llogd_body *d);
1901 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
1902 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
1903 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
1904 struct llog_rec_tail *tail);
1907 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
1909 /* quota. fixed by tianzy for bug10707 */
1910 #define QUOTA_IS_GRP 0X1UL /* 0 is user, 1 is group. Used by qd_flags*/
1911 #define QUOTA_IS_BLOCK 0x2UL /* 0 is inode, 1 is block. Used by qd_flags*/
1914 __u32 qd_id; /* ID appiles to (uid, gid) */
1915 __u32 qd_flags; /* Quota type (USRQUOTA, GRPQUOTA) occupy one bit;
1916 * Block quota or file quota occupy one bit */
1917 __u64 qd_count; /* acquire/release count (bytes for block quota) */
1920 struct qunit_data_old {
1921 __u32 qd_id; /* ID appiles to (uid, gid) */
1922 __u32 qd_type; /* Quota type (USRQUOTA, GRPQUOTA) */
1923 __u32 qd_count; /* acquire/release count (bytes for block quota) */
1924 __u32 qd_isblk; /* Block quota or file quota */
1927 extern void lustre_swab_qdata(struct qunit_data *d);
1928 extern void lustre_swab_qdata_old(struct qunit_data_old *d);
1929 extern struct qunit_data *lustre_quota_old_to_new(struct qunit_data_old *d);
1930 extern struct qunit_data_old *lustre_quota_new_to_old(struct qunit_data *d);
1937 #define JOIN_FILE_ALIGN 4096
1939 /* security opcodes */
1942 SEC_CTX_INIT_CONT = 802,
1948 * capa related definitions
1950 #define CAPA_HMAC_MAX_LEN 64
1951 #define CAPA_HMAC_KEY_MAX_LEN 56
1953 /* NB take care when changing the sequence of elements this struct,
1954 * because the offset info is used in find_capa() */
1955 struct lustre_capa {
1956 struct lu_fid lc_fid; /* fid */
1957 __u64 lc_opc; /* operations allowed */
1958 __u32 lc_flags; /* HMAC algorithm & flags */
1959 __u32 lc_keyid; /* key used for the capability */
1960 __u64 lc_expiry; /* expiry time (sec) */
1961 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /* HMAC */
1962 } __attribute__((packed));
1964 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
1966 /* lustre_capa.lc_opc */
1968 CAPA_OPC_BODY_WRITE = 1<<0, /* write object data */
1969 CAPA_OPC_BODY_READ = 1<<1, /* read object data */
1970 CAPA_OPC_INDEX_LOOKUP = 1<<2, /* lookup object fid */
1971 CAPA_OPC_INDEX_INSERT = 1<<3, /* insert object fid */
1972 CAPA_OPC_INDEX_DELETE = 1<<4, /* delete object fid */
1973 CAPA_OPC_OSS_WRITE = 1<<5, /* write oss object data */
1974 CAPA_OPC_OSS_READ = 1<<6, /* read oss object data */
1975 CAPA_OPC_OSS_TRUNC = 1<<7, /* truncate oss object */
1976 CAPA_OPC_META_WRITE = 1<<8, /* write object meta data */
1977 CAPA_OPC_META_READ = 1<<9, /* read object meta data */
1981 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
1982 #define CAPA_OPC_MDS_ONLY \
1983 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
1984 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
1985 #define CAPA_OPC_OSS_ONLY \
1986 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC)
1987 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
1988 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
1990 /* MDS capability covers object capability for operations of body r/w
1991 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
1992 * while OSS capability only covers object capability for operations of
1993 * oss data(file content) r/w/truncate.
1995 static inline int capa_for_mds(struct lustre_capa *c)
1997 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2000 static inline int capa_for_oss(struct lustre_capa *c)
2002 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2005 /* lustre_capa.lc_flags */
2007 CAPA_FL_SHORT_EXPIRY = 1, /* short capa expiry */
2010 /* lustre_capa.lc_hmac_alg */
2012 CAPA_HMAC_ALG_SHA1 = 1, /* sha1 algorithm */
2016 #define CAPA_FL_MASK 0x00ffffff
2017 #define CAPA_HMAC_ALG_MASK 0xff000000
2019 struct lustre_capa_key {
2020 __u64 lk_mdsid; /* mds# */
2021 __u32 lk_keyid; /* key# */
2023 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /* key */
2024 } __attribute__((packed));
2026 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2028 typedef int (* renew_capa_cb_t)(struct obd_capa *, struct lustre_capa *);