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 LASSERT(fid_is_igif(f0) || fid_ver(f0) == 0);
273 LASSERT(fid_is_igif(f1) || fid_ver(f1) == 0);
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 MEA_MAGIC_LAST_CHAR 0xb2221ca1
330 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
331 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
333 #define MAX_HASH_SIZE 0x7fffffffUL
334 #define MAX_HASH_HIGHEST_BIT 0x10000000
336 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
337 struct lmv_stripe_md {
341 struct lu_fid mea_ids[0];
344 struct lustre_handle {
347 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
349 static inline int lustre_handle_is_used(struct lustre_handle *lh)
351 return lh->cookie != 0ull;
354 static inline int lustre_handle_equal(struct lustre_handle *lh1,
355 struct lustre_handle *lh2)
357 return lh1->cookie == lh2->cookie;
360 static inline void lustre_handle_copy(struct lustre_handle *tgt,
361 struct lustre_handle *src)
363 tgt->cookie = src->cookie;
366 /* we depend on this structure to be 8-byte aligned */
367 /* this type is only endian-adjusted in lustre_unpack_msg() */
368 struct lustre_msg_v1 {
369 struct lustre_handle lm_handle;
375 __u64 lm_last_committed;
384 #define lustre_msg lustre_msg_v2
385 /* we depend on this structure to be 8-byte aligned */
386 /* this type is only endian-adjusted in lustre_unpack_msg() */
387 struct lustre_msg_v2 {
395 /* without gss, ptlrpc_body is put at the first buffer. */
397 struct lustre_handle pb_handle;
403 __u64 pb_last_committed;
408 __u32 pb_paddings[3];
411 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
413 /* message body offset for lustre_msg_v2 */
414 /* ptlrpc body offset in all request/reply messages */
415 #define MSG_PTLRPC_BODY_OFF 0
417 /* normal request/reply message record offset */
418 #define REQ_REC_OFF 1
419 #define REPLY_REC_OFF 1
421 /* ldlm request message body offset */
422 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
423 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
425 /* ldlm intent lock message body offset */
426 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
427 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
429 /* ldlm reply message body offset */
430 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
431 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
433 /* Flags that are operation-specific go in the top 16 bits. */
434 #define MSG_OP_FLAG_MASK 0xffff0000
435 #define MSG_OP_FLAG_SHIFT 16
437 /* Flags that apply to all requests are in the bottom 16 bits */
438 #define MSG_GEN_FLAG_MASK 0x0000ffff
439 #define MSG_LAST_REPLAY 1
442 #define MSG_REQ_REPLAY_DONE 8
443 #define MSG_LOCK_REPLAY_DONE 16
446 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
449 #define MSG_CONNECT_RECOVERING 0x1
450 #define MSG_CONNECT_RECONNECT 0x2
451 #define MSG_CONNECT_REPLAYABLE 0x4
452 //#define MSG_CONNECT_PEER 0x8
453 #define MSG_CONNECT_LIBCLIENT 0x10
454 #define MSG_CONNECT_INITIAL 0x20
455 #define MSG_CONNECT_ASYNC 0x40
456 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
457 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
460 #define OBD_CONNECT_RDONLY 0x1ULL /* client allowed read-only access */
461 #define OBD_CONNECT_INDEX 0x2ULL /* connect to specific LOV idx */
462 #define OBD_CONNECT_GRANT 0x8ULL /* OSC acquires grant at connect */
463 #define OBD_CONNECT_SRVLOCK 0x10ULL /* server takes locks for client */
464 #define OBD_CONNECT_VERSION 0x20ULL /* Server supports versions in ocd */
465 #define OBD_CONNECT_REQPORTAL 0x40ULL /* Separate portal for non-IO reqs */
466 #define OBD_CONNECT_ACL 0x80ULL /* client uses access control lists */
467 #define OBD_CONNECT_XATTR 0x100ULL /* client using extended attributes*/
468 #define OBD_CONNECT_CROW 0x200ULL /* MDS+OST create objects on write */
469 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /* locks on server for punch b=9528 */
470 #define OBD_CONNECT_TRANSNO 0x800ULL /* replay sends initial transno */
471 #define OBD_CONNECT_IBITS 0x1000ULL /* support for inodebits locks */
472 #define OBD_CONNECT_JOIN 0x2000ULL /* files can be concatenated */
473 #define OBD_CONNECT_REAL 0x4000ULL
474 #define OBD_CONNECT_ATTRFID 0x8000ULL /* Server supports GetAttr By Fid */
475 #define OBD_CONNECT_NODEVOH 0x10000ULL /* No open handle for special nodes */
476 #define OBD_CONNECT_LCL_CLIENT 0x20000ULL /* local 1.8 client */
477 #define OBD_CONNECT_RMT_CLIENT 0x40000ULL /* Remote 1.8 client */
478 #define OBD_CONNECT_BRW_SIZE 0x80000ULL /* Max bytes per rpc */
479 #define OBD_CONNECT_QUOTA64 0x100000ULL /* 64bit qunit_data.qd_count b=10707*/
480 #define OBD_CONNECT_MDS_CAPA 0x200000ULL /* MDS capability */
481 #define OBD_CONNECT_OSS_CAPA 0x400000ULL /* OSS capability */
482 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
483 * and lustre/utils/wirecheck.c */
485 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
486 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
487 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
488 OBD_CONNECT_NODEVOH | OBD_CONNECT_ATTRFID | \
489 OBD_CONNECT_LCL_CLIENT | \
490 OBD_CONNECT_RMT_CLIENT | \
491 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA)
492 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
493 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
494 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
495 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
496 OBD_CONNECT_OSS_CAPA)
497 #define ECHO_CONNECT_SUPPORTED (0)
498 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION)
500 #define MAX_QUOTA_COUNT32 ((0xffffffffULL >> QUOTABLOCK_BITS) << QUOTABLOCK_BITS)
502 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
503 ((patch)<<8) + (fix))
504 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
505 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
506 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
507 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
509 /* This structure is used for both request and reply.
511 * If we eventually have separate connect data for different types, which we
512 * almost certainly will, then perhaps we stick a union in here. */
513 struct obd_connect_data {
514 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
515 __u64 ocd_transno; /* first transno from client to be replayed */
516 __u64 ocd_ibits_known; /* inode bits this client understands */
517 __u32 ocd_version; /* lustre release version number */
518 __u32 ocd_grant; /* initial cache grant amount (bytes) */
519 __u32 ocd_index; /* LOV index to connect to */
520 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
521 __u32 ocd_nllu; /* non-local-lustre-user */
522 __u32 ocd_nllg; /* non-local-lustre-group */
523 __u32 ocd_group; /* MDS group on OST */
524 __u32 padding1; /* also fix lustre_swab_connect */
525 __u64 padding2; /* also fix lustre_swab_connect */
526 __u64 padding3; /* also fix lustre_swab_connect */
529 extern void lustre_swab_connect(struct obd_connect_data *ocd);
532 * OST requests: OBDO & OBD request records
537 OST_REPLY = 0, /* reply ? */
551 /* OST_SAN_READ = 14, deprecated */
552 /* OST_SAN_WRITE = 15, deprecated */
559 #define OST_FIRST_OPC OST_REPLY
561 typedef uint64_t obd_id;
562 typedef uint64_t obd_gr;
563 typedef uint64_t obd_time;
564 typedef uint64_t obd_size;
565 typedef uint64_t obd_off;
566 typedef uint64_t obd_blocks;
567 typedef uint32_t obd_blksize;
568 typedef uint32_t obd_mode;
569 typedef uint32_t obd_uid;
570 typedef uint32_t obd_gid;
571 typedef uint32_t obd_flag;
572 typedef uint64_t obd_valid;
573 typedef uint32_t obd_count;
575 #define OBD_FL_INLINEDATA (0x00000001)
576 #define OBD_FL_OBDMDEXISTS (0x00000002)
577 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
578 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
579 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
580 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
581 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
582 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
583 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
584 #define OBD_FL_CREATE_CROW (0x00000400) /* object should be create on write */
587 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
588 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
591 #define OBD_FL_TRUNCLOCK (0x00000800)
594 * This should not be smaller than sizeof(struct lustre_handle) + sizeof(struct
595 * llog_cookie) + sizeof(struct ll_fid). Nevertheless struct ll_fid is not
596 * longer stored in o_inline, we keep this just for case.
598 #define OBD_INLINESZ 80
600 /* Note: 64-bit types are 64-bit aligned in structure */
602 obd_valid o_valid; /* hot fields in this obdo */
606 obd_size o_size; /* o_size-o_blocks == ost_lvb */
610 obd_blocks o_blocks; /* brw: cli sent cached bytes */
613 /* 32-bit fields start here: keep an even number of them via padding */
614 obd_blksize o_blksize; /* optimal IO blocksize */
615 obd_mode o_mode; /* brw: cli sent cache remain */
619 obd_count o_nlink; /* brw: checksum */
620 obd_count o_generation;
621 obd_count o_misc; /* brw: o_dropped */
622 __u32 o_easize; /* epoch in ost writes */
624 __u32 o_stripe_idx; /* holds stripe idx */
626 char o_inline[OBD_INLINESZ];
627 /* lustre_handle + llog_cookie */
630 #define o_dirty o_blocks
631 #define o_undirty o_mode
632 #define o_dropped o_misc
633 #define o_cksum o_nlink
635 extern void lustre_swab_obdo (struct obdo *o);
638 struct lu_fid op_fid1;
639 struct lu_fid op_fid2;
640 struct lustre_handle op_handle;
645 struct lmv_stripe_md *op_mea1;
646 struct lmv_stripe_md *op_mea2;
647 __u32 op_suppgids[2];
652 /* iattr fields and blocks. */
653 struct iattr op_attr;
655 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,14)
656 unsigned int op_attr_flags;
659 loff_t op_attr_blocks;
661 /* Size-on-MDS epoch and flags. */
666 struct obd_capa *op_capa1;
667 struct obd_capa *op_capa2;
669 /* Should server check split in lookups or not. */
673 #define MDS_MODE_DONT_LOCK (1 << 30)
674 #define MDS_MODE_REPLAY (1 << 31)
676 #define LOV_MAGIC_V1 0x0BD10BD0
677 #define LOV_MAGIC LOV_MAGIC_V1
678 #define LOV_MAGIC_JOIN 0x0BD20BD0
680 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
681 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
682 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
683 #define LOV_PATTERN_CMOBD 0x200
685 #define lov_ost_data lov_ost_data_v1
686 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
687 __u64 l_object_id; /* OST object ID */
688 __u64 l_object_gr; /* OST object group (creating MDS number) */
689 __u32 l_ost_gen; /* generation of this l_ost_idx */
690 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
693 #define lov_mds_md lov_mds_md_v1
694 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
695 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
696 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
697 __u64 lmm_object_id; /* LOV object ID */
698 __u64 lmm_object_gr; /* LOV object group */
699 __u32 lmm_stripe_size; /* size of stripe in bytes */
700 __u32 lmm_stripe_count; /* num stripes in use for this object */
701 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
704 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
705 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
707 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
708 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
709 #define XATTR_NAME_LOV "trusted.lov"
712 #define XATTR_NAME_LUSTRE_ACL "system.lustre_acl"
714 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
715 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
716 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
717 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
718 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
719 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
720 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
721 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
722 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
723 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
724 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
725 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
726 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
727 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
728 #define OBD_MD_FLINLINE (0x00008000ULL) /* inline data */
729 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
730 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
731 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
732 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file handle */
733 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
734 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
735 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
736 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
737 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
738 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
739 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
740 /* ->mds if epoch opens or closes */
741 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
742 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
743 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
744 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
745 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
747 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
748 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
749 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD EA */
751 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
752 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
753 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
754 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
755 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
756 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
757 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
758 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
760 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
761 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
762 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
763 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
764 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
766 static inline struct lustre_handle *obdo_handle(struct obdo *oa)
768 return (struct lustre_handle *)oa->o_inline;
771 static inline struct llog_cookie *obdo_logcookie(struct obdo *oa)
773 return (struct llog_cookie *)(oa->o_inline +
774 sizeof(struct lustre_handle));
776 /* don't forget obdo_fid which is way down at the bottom so it can
777 * come after the definition of llog_cookie */
790 __u32 os_state; /* positive error code on server */
802 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
804 /* ost_body.data values for OST_BRW */
806 #define OBD_BRW_READ 0x01
807 #define OBD_BRW_WRITE 0x02
808 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
809 #define OBD_BRW_SYNC 0x08
810 #define OBD_BRW_CHECK 0x10
811 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
812 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
813 #define OBD_BRW_DROP 0x80 /* drop the page after IO */
814 #define OBD_BRW_NOQUOTA 0x100
815 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
817 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
819 #define OST_MIN_PRECREATE 16384
820 #define OST_MAX_PRECREATE 20000
829 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
831 /* multiple of 8 bytes => can array */
832 struct niobuf_remote {
838 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
840 /* request structure for OST's */
846 extern void lustre_swab_ost_body (struct ost_body *b);
847 extern void lustre_swab_ost_last_id(obd_id *id);
848 extern void lustre_swab_generic_32s(__u32 *val);
850 /* lock value block communicated between the filter and llite */
860 extern void lustre_swab_ost_lvb(struct ost_lvb *);
869 MDS_GETATTR_NAME = 34,
880 MDS_DONE_WRITING = 45,
891 #define MDS_FIRST_OPC MDS_GETATTR
907 } mds_reint_t, mdt_reint_t;
909 /* the disposition of the intent outlines what was executed */
910 #define DISP_IT_EXECD 0x00000001
911 #define DISP_LOOKUP_EXECD 0x00000002
912 #define DISP_LOOKUP_NEG 0x00000004
913 #define DISP_LOOKUP_POS 0x00000008
914 #define DISP_OPEN_CREATE 0x00000010
915 #define DISP_OPEN_OPEN 0x00000020
916 #define DISP_ENQ_COMPLETE 0x00400000
917 #define DISP_ENQ_OPEN_REF 0x00800000
918 #define DISP_ENQ_CREATE_REF 0x01000000
920 /* INODE LOCK PARTS */
921 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
922 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
923 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
925 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
926 #define MDS_INODELOCK_MAXSHIFT 2
927 /* This FULL lock is useful to take on unlink sort of operations */
928 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
930 #define LUSTRE_CONFIG_SET 0
931 #define LUSTRE_CONFIG_GET 1
933 #define LUSTRE_CONFIG_METASEQ "metaseq"
934 #define LUSTRE_CONFIG_TRANSNO "transno"
937 __u64 id; /* holds object id */
938 __u32 generation; /* holds object generation */
940 __u32 f_type; /* holds object type or stripe idx when passing it to
941 * OST for saving into EA. */
944 extern void lustre_swab_ll_fid (struct ll_fid *fid);
946 #define MDS_STATUS_CONN 1
947 #define MDS_STATUS_LOV 2
949 struct mds_status_req {
954 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
956 /* mdt_thread_info.mti_flags. */
957 enum mdt_ioepoch_flags {
958 /* The flag indicates Size-on-MDS attributes are changed. */
959 MF_SOM_CHANGE = (1 << 0),
960 /* Flags indicates an epoch opens or closes. */
961 MF_EPOCH_OPEN = (1 << 1),
962 MF_EPOCH_CLOSE = (1 << 2),
965 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
967 /* these should be identical to their EXT3_*_FL counterparts, and are
968 * redefined here only to avoid dragging in ext3_fs.h */
969 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
970 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
971 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
972 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
973 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
976 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
977 * and we need to decode these into local S_* flags in the inode. Otherwise
978 * we pass flags straight through (see bug 9486). */
979 static inline int ll_ext_to_inode_flags(int flags)
981 return (flags & MDS_BFLAG_EXT_FLAGS) ?
982 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
983 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
984 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
985 #if defined(S_DIRSYNC)
986 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
988 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
989 (flags & ~MDS_BFLAG_EXT_FLAGS);
992 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
993 * and we pass these straight through. Otherwise we need to convert from
994 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
995 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
997 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
998 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
999 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1000 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1001 #if defined(S_DIRSYNC)
1002 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1004 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1011 struct lustre_handle handle;
1013 __u64 size; /* Offset, in the case of MDS_READPAGE */
1017 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1025 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1027 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1032 __u32 max_cookiesize; /* also fix lustre_swab_mdt_body */
1038 struct lustre_handle handle;
1040 __u64 size; /* Offset, in the case of MDS_READPAGE */
1044 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1053 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1055 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1061 __u32 max_cookiesize; /* also fix lustre_swab_mds_body */
1062 __u32 padding_4; /* also fix lustre_swab_mds_body */
1065 extern void lustre_swab_mds_body (struct mds_body *b);
1066 extern void lustre_swab_mdt_body (struct mdt_body *b);
1069 struct lustre_handle handle;
1074 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1077 struct mdt_body *body;
1078 struct lov_stripe_md *lsm;
1079 struct lmv_stripe_md *mea;
1080 #ifdef CONFIG_FS_POSIX_ACL
1081 struct posix_acl *posix_acl;
1083 struct mdt_remote_perm *remote_perm;
1084 struct obd_capa *mds_capa;
1085 struct obd_capa *oss_capa;
1088 #define Q_QUOTACHECK 0x800100
1089 #define Q_INITQUOTA 0x800101 /* init slave limits */
1090 #define Q_GETOINFO 0x800102 /* get obd quota info */
1091 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1093 #define Q_TYPESET(oqc, type) \
1094 ((oqc)->qc_type == type || (oqc)->qc_type == UGQUOTA)
1096 #define Q_GETOCMD(oqc) \
1097 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1099 struct obd_quotactl {
1104 struct obd_dqinfo qc_dqinfo;
1105 struct obd_dqblk qc_dqblk;
1108 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1110 /* inode access permission for remote user, the inode info are omitted,
1111 * for client knows them. */
1112 struct mds_remote_perm {
1117 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1120 /* setxid permissions for mds_setxid_perm.mp_perm */
1121 #define LUSTRE_SETUID_PERM 0x01
1122 #define LUSTRE_SETGID_PERM 0x02
1123 #define LUSTRE_SETGRP_PERM 0x04
1125 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1127 struct mdt_remote_perm {
1132 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1135 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1137 struct mds_rec_setattr {
1144 struct ll_fid sa_fid;
1152 __u32 sa_attr_flags;
1153 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1156 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1158 struct mdt_rec_setattr {
1165 struct lu_fid sa_fid;
1174 __u32 sa_attr_flags;
1175 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1178 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1180 /* Remove this once we declare it in include/linux/fs.h (v21 kernel patch?) */
1181 #ifndef ATTR_CTIME_SET
1182 #define ATTR_CTIME_SET 0x2000
1186 #define FMODE_READ 00000001
1187 #define FMODE_WRITE 00000002
1190 #define FMODE_EPOCH 01000000
1191 #define FMODE_EPOCHLCK 02000000
1192 #define FMODE_SOM 04000000
1193 #define FMODE_CLOSED 0
1195 #define MDS_OPEN_CREATED 00000010
1197 #define MDS_FMODE_EXEC 00000004
1198 #define MDS_OPEN_CREAT 00000100
1199 #define MDS_OPEN_EXCL 00000200
1200 #define MDS_OPEN_TRUNC 00001000
1201 #define MDS_OPEN_APPEND 00002000
1202 #define MDS_OPEN_SYNC 00010000
1203 #define MDS_OPEN_DIRECTORY 00200000
1205 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1206 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1207 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1208 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1209 * with default ACL */
1210 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1211 * actually, this is for create, not
1212 * conflict with other open flags */
1213 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1214 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1215 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1217 struct mds_rec_join {
1218 struct ll_fid jr_fid;
1222 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1224 struct mdt_rec_join {
1225 struct lu_fid jr_fid;
1229 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1231 struct mds_rec_create {
1236 __u32 cr_flags; /* for use with open */
1238 struct ll_fid cr_fid;
1239 struct ll_fid cr_replayfid;
1243 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1244 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1245 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1246 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1247 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1250 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1252 struct mdt_rec_create {
1257 __u32 cr_flags; /* for use with open */
1259 struct lu_fid cr_fid1;
1260 struct lu_fid cr_fid2;
1266 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1267 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1270 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1272 struct mds_rec_link {
1279 struct ll_fid lk_fid1;
1280 struct ll_fid lk_fid2;
1282 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1283 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1284 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1285 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1288 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1290 struct mdt_rec_link {
1297 struct lu_fid lk_fid1;
1298 struct lu_fid lk_fid2;
1301 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1302 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1303 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1306 extern void lustre_swab_mdt_rec_link (struct mdt_rec_link *lk);
1308 struct mds_rec_unlink {
1315 struct ll_fid ul_fid1;
1316 struct ll_fid ul_fid2;
1318 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1319 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1320 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1321 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1324 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1326 struct mdt_rec_unlink {
1333 struct lu_fid ul_fid1;
1334 struct lu_fid ul_fid2;
1337 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1338 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1339 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1342 extern void lustre_swab_mdt_rec_unlink (struct mdt_rec_unlink *ul);
1344 struct mds_rec_rename {
1351 struct ll_fid rn_fid1;
1352 struct ll_fid rn_fid2;
1354 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1355 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1356 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1357 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1360 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1362 struct mdt_rec_rename {
1369 struct lu_fid rn_fid1;
1370 struct lu_fid rn_fid2;
1372 __u32 rn_mode; /* cross-ref rename has mode */
1373 __u32 rn_cksplit; /* check for split or not */
1374 __u32 rn_padding_3; /* also fix lustre_swab_mdt_rec_rename */
1375 __u32 rn_padding_4; /* also fix lustre_swab_mdt_rec_rename */
1378 extern void lustre_swab_mdt_rec_rename (struct mdt_rec_rename *rn);
1380 /* begin adding MDT by huanghua@clusterfs.com */
1382 __u32 ld_tgt_count; /* how many MDS's */
1383 __u32 ld_active_tgt_count; /* how many active */
1384 struct obd_uuid ld_uuid;
1387 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1388 /* end adding MDT by huanghua@clusterfs.com */
1395 extern void lustre_swab_md_fld (struct md_fld *mf);
1400 FLD_FIRST_OPC = FLD_QUERY
1406 SEQ_FIRST_OPC = SEQ_QUERY
1410 SEQ_ALLOC_SUPER = 0,
1415 * LOV data structures
1418 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1419 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1421 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1422 /* The size of the buffer the lov/mdc reserves for the
1423 * array of UUIDs returned by the MDS. With the current
1424 * protocol, this will limit the max number of OSTs per LOV */
1426 #define LOV_DESC_MAGIC 0xB0CCDE5C
1428 /* LOV settings descriptor (should only contain static info) */
1430 __u32 ld_tgt_count; /* how many OBD's */
1431 __u32 ld_active_tgt_count; /* how many active */
1432 __u32 ld_default_stripe_count; /* how many objects are used */
1433 __u32 ld_pattern; /* default PATTERN_RAID0 */
1434 __u64 ld_default_stripe_size; /* in bytes */
1435 __u64 ld_default_stripe_offset; /* in bytes */
1436 __u32 ld_padding_0; /* unused */
1437 __u32 ld_qos_maxage; /* in second */
1438 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1439 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1440 struct obd_uuid ld_uuid;
1443 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1445 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1450 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1455 LDLM_BL_CALLBACK = 104,
1456 LDLM_CP_CALLBACK = 105,
1457 LDLM_GL_CALLBACK = 106,
1460 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1462 #define RES_NAME_SIZE 4
1463 struct ldlm_res_id {
1464 __u64 name[RES_NAME_SIZE];
1467 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1490 #define LDLM_MIN_TYPE LDLM_PLAIN
1492 struct ldlm_extent {
1498 struct ldlm_inodebits {
1505 __u64 blocking_export; /* not actually used over the wire */
1506 __u32 blocking_pid; /* not actually used over the wire */
1510 /* it's important that the fields of the ldlm_extent structure match
1511 * the first fields of the ldlm_flock structure because there is only
1512 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1513 * this ever changes we will need to swab the union differently based
1514 * on the resource type. */
1517 struct ldlm_extent l_extent;
1518 struct ldlm_flock l_flock;
1519 struct ldlm_inodebits l_inodebits;
1520 } ldlm_policy_data_t;
1522 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1524 struct ldlm_intent {
1528 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1530 struct ldlm_resource_desc {
1531 ldlm_type_t lr_type;
1532 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1533 struct ldlm_res_id lr_name;
1536 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1538 struct ldlm_lock_desc {
1539 struct ldlm_resource_desc l_resource;
1540 ldlm_mode_t l_req_mode;
1541 ldlm_mode_t l_granted_mode;
1542 ldlm_policy_data_t l_policy_data;
1545 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1547 struct ldlm_request {
1549 __u32 lock_padding; /* also fix lustre_swab_ldlm_request */
1550 struct ldlm_lock_desc lock_desc;
1551 struct lustre_handle lock_handle1;
1552 struct lustre_handle lock_handle2;
1555 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1559 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1560 struct ldlm_lock_desc lock_desc;
1561 struct lustre_handle lock_handle;
1562 __u64 lock_policy_res1;
1563 __u64 lock_policy_res2;
1566 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1570 * Opcodes for mountconf (mgs and mgc)
1575 MGS_EXCEPTION, /* node died, etc. */
1576 MGS_TARGET_REG, /* whenever target starts up */
1581 /* We pass this info to the MGS so it can write config logs */
1582 #define MTI_NAME_MAXLEN 64
1583 #define MTI_NIDS_MAX 32
1584 struct mgs_target_info {
1585 __u32 mti_lustre_ver;
1586 __u32 mti_stripe_index;
1587 __u32 mti_config_ver;
1589 __u32 mti_nid_count;
1590 __u32 padding; /* 64 bit align */
1591 char mti_fsname[MTI_NAME_MAXLEN];
1592 char mti_svname[MTI_NAME_MAXLEN];
1593 char mti_uuid[sizeof(struct obd_uuid)];
1594 lnet_nid_t mti_nids[MTI_NIDS_MAX]; /* host nids */
1595 char mti_params[2048];
1598 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
1600 #define CM_START 0x01
1602 #define CM_SKIP 0x04
1603 #define CM_UPGRADE146 0x08
1604 #define CM_START_SKIP (CM_START | CM_SKIP)
1607 __u32 cm_step; /* aka config version */
1609 __u32 cm_vers; /* lustre release version number */
1610 __u32 padding; /* 64 bit align */
1611 time_t cm_createtime; /*when this record was first created */
1612 time_t cm_canceltime; /*when this record is no longer valid*/
1613 char cm_svname[MTI_NAME_MAXLEN];
1614 char cm_comment[MTI_NAME_MAXLEN];
1618 * Opcodes for multiple servers.
1627 #define OBD_FIRST_OPC OBD_PING
1629 /* catalog of log objects */
1631 /* Identifier for a single log object */
1636 } __attribute__((packed));
1638 /* Records written to the CATALOGS list */
1639 #define CATLIST "CATALOGS"
1641 struct llog_logid lci_logid;
1645 } __attribute__((packed));
1647 /*join file lov mds md*/
1648 struct lov_mds_md_join {
1649 struct lov_mds_md lmmj_md;
1650 /*join private info*/
1651 struct llog_logid lmmj_array_id; /*array object id*/
1652 __u32 lmmj_extent_count; /*array extent count*/
1655 /* Log data record types - there is no specific reason that these need to
1656 * be related to the RPC opcodes, but no reason not to (may be handy later?)
1658 #define LLOG_OP_MAGIC 0x10600000
1659 #define LLOG_OP_MASK 0xfff00000
1662 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
1663 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
1664 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
1665 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
1666 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
1667 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
1668 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
1669 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
1670 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
1671 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
1672 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
1676 * for now, continue to support old pad records which have 0 for their
1677 * type but still need to be swabbed for their length
1679 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
1680 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
1681 __swab32(LLOG_OP_MAGIC) || \
1682 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
1684 /* Log record header - stored in little endian order.
1685 * Each record must start with this struct, end with a llog_rec_tail,
1686 * and be a multiple of 256 bits in size.
1688 struct llog_rec_hdr {
1695 struct llog_rec_tail {
1700 struct llog_logid_rec {
1701 struct llog_rec_hdr lid_hdr;
1702 struct llog_logid lid_id;
1708 struct llog_rec_tail lid_tail;
1709 } __attribute__((packed));
1711 /* MDS extent description
1712 * It is for joined file extent info, each extent info for joined file
1713 * just like (start, end, lmm).
1715 struct mds_extent_desc {
1716 __u64 med_start; /* extent start */
1717 __u64 med_len; /* extent length */
1718 struct lov_mds_md med_lmm; /* extent's lmm */
1720 /*Joined file array extent log record*/
1721 struct llog_array_rec {
1722 struct llog_rec_hdr lmr_hdr;
1723 struct mds_extent_desc lmr_med;
1724 struct llog_rec_tail lmr_tail;
1727 struct llog_create_rec {
1728 struct llog_rec_hdr lcr_hdr;
1729 struct ll_fid lcr_fid;
1733 struct llog_rec_tail lcr_tail;
1734 } __attribute__((packed));
1736 struct llog_orphan_rec {
1737 struct llog_rec_hdr lor_hdr;
1741 struct llog_rec_tail lor_tail;
1742 } __attribute__((packed));
1744 struct llog_unlink_rec {
1745 struct llog_rec_hdr lur_hdr;
1749 struct llog_rec_tail lur_tail;
1750 } __attribute__((packed));
1752 struct llog_setattr_rec {
1753 struct llog_rec_hdr lsr_hdr;
1759 struct llog_rec_tail lsr_tail;
1760 } __attribute__((packed));
1762 struct llog_size_change_rec {
1763 struct llog_rec_hdr lsc_hdr;
1764 struct ll_fid lsc_fid;
1767 struct llog_rec_tail lsc_tail;
1768 } __attribute__((packed));
1773 } __attribute__((packed));
1775 struct llog_gen_rec {
1776 struct llog_rec_hdr lgr_hdr;
1777 struct llog_gen lgr_gen;
1778 struct llog_rec_tail lgr_tail;
1780 /* On-disk header structure of each log object, stored in little endian order */
1781 #define LLOG_CHUNK_SIZE 8192
1782 #define LLOG_HEADER_SIZE (96)
1783 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
1785 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
1787 /* flags for the logs */
1788 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
1789 #define LLOG_F_IS_CAT 0x2
1790 #define LLOG_F_IS_PLAIN 0x4
1792 struct llog_log_hdr {
1793 struct llog_rec_hdr llh_hdr;
1794 __u64 llh_timestamp;
1796 __u32 llh_bitmap_offset;
1800 /* for a catalog the first plain slot is next to it */
1801 struct obd_uuid llh_tgtuuid;
1802 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
1803 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
1804 struct llog_rec_tail llh_tail;
1805 } __attribute__((packed));
1807 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
1808 llh->llh_bitmap_offset - \
1809 sizeof(llh->llh_tail)) * 8)
1811 /* log cookies are used to reference a specific log file and a record therein */
1812 struct llog_cookie {
1813 struct llog_logid lgc_lgl;
1817 } __attribute__((packed));
1820 enum llogd_rpc_ops {
1821 LLOG_ORIGIN_HANDLE_CREATE = 501,
1822 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
1823 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
1824 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
1825 LLOG_ORIGIN_HANDLE_CLOSE = 505,
1826 LLOG_ORIGIN_CONNECT = 506,
1827 LLOG_CATINFO = 507, /* for lfs catinfo */
1828 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
1829 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
1834 struct llog_logid lgd_logid;
1836 __u32 lgd_llh_flags;
1838 __u32 lgd_saved_index;
1840 __u64 lgd_cur_offset;
1841 } __attribute__((packed));
1843 struct llogd_conn_body {
1844 struct llog_gen lgdc_gen;
1845 struct llog_logid lgdc_logid;
1846 __u32 lgdc_ctxt_idx;
1847 } __attribute__((packed));
1849 struct lov_user_ost_data_join { /* per-stripe data structure */
1850 __u64 l_extent_start; /* extent start*/
1851 __u64 l_extent_end; /* extent end*/
1852 __u64 l_object_id; /* OST object ID */
1853 __u64 l_object_gr; /* OST object group (creating MDS number) */
1854 __u32 l_ost_gen; /* generation of this OST index */
1855 __u32 l_ost_idx; /* OST index in LOV */
1856 } __attribute__((packed));
1858 struct lov_user_md_join { /* LOV EA user data (host-endian) */
1859 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
1860 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1861 __u64 lmm_object_id; /* LOV object ID */
1862 __u64 lmm_object_gr; /* LOV object group */
1863 __u32 lmm_stripe_size; /* size of stripe in bytes */
1864 __u32 lmm_stripe_count; /* num stripes in use for this object */
1865 __u32 lmm_extent_count; /* extent count of lmm*/
1866 __u64 lmm_tree_id; /* mds tree object id */
1867 __u64 lmm_tree_gen; /* mds tree object gen */
1868 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
1869 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
1870 } __attribute__((packed));
1872 extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm);
1873 extern void lustre_swab_lov_user_md(struct lov_user_md *lum);
1874 extern void lustre_swab_lov_user_md_objects(struct lov_user_md *lum);
1875 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
1878 extern void lustre_swab_llogd_body (struct llogd_body *d);
1879 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
1880 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
1881 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
1882 struct llog_rec_tail *tail);
1885 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
1887 /* quota. fixed by tianzy for bug10707 */
1888 #define QUOTA_IS_GRP 0X1UL /* 0 is user, 1 is group. Used by qd_flags*/
1889 #define QUOTA_IS_BLOCK 0x2UL /* 0 is inode, 1 is block. Used by qd_flags*/
1892 __u32 qd_id; /* ID appiles to (uid, gid) */
1893 __u32 qd_flags; /* Quota type (USRQUOTA, GRPQUOTA) occupy one bit;
1894 * Block quota or file quota occupy one bit */
1895 __u64 qd_count; /* acquire/release count (bytes for block quota) */
1898 struct qunit_data_old {
1899 __u32 qd_id; /* ID appiles to (uid, gid) */
1900 __u32 qd_type; /* Quota type (USRQUOTA, GRPQUOTA) */
1901 __u32 qd_count; /* acquire/release count (bytes for block quota) */
1902 __u32 qd_isblk; /* Block quota or file quota */
1905 extern void lustre_swab_qdata(struct qunit_data *d);
1906 extern void lustre_swab_qdata_old(struct qunit_data_old *d);
1907 extern struct qunit_data *lustre_quota_old_to_new(struct qunit_data_old *d);
1908 extern struct qunit_data_old *lustre_quota_new_to_old(struct qunit_data *d);
1915 #define JOIN_FILE_ALIGN 4096
1917 /* security opcodes */
1920 SEC_CTX_INIT_CONT = 802,
1926 * capa related definitions
1928 #define CAPA_HMAC_MAX_LEN 64
1929 #define CAPA_HMAC_KEY_MAX_LEN 56
1931 /* NB take care when changing the sequence of elements this struct,
1932 * because the offset info is used in find_capa() */
1933 struct lustre_capa {
1934 struct lu_fid lc_fid; /* fid */
1935 __u64 lc_opc; /* operations allowed */
1936 __u32 lc_flags; /* HMAC algorithm & flags */
1937 __u32 lc_keyid; /* key used for the capability */
1938 __u64 lc_expiry; /* expiry time (sec) */
1939 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /* HMAC */
1940 } __attribute__((packed));
1942 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
1944 /* lustre_capa.lc_opc */
1946 CAPA_OPC_BODY_WRITE = 1<<0, /* write object data */
1947 CAPA_OPC_BODY_READ = 1<<1, /* read object data */
1948 CAPA_OPC_INDEX_LOOKUP = 1<<2, /* lookup object fid */
1949 CAPA_OPC_INDEX_INSERT = 1<<3, /* insert object fid */
1950 CAPA_OPC_INDEX_DELETE = 1<<4, /* delete object fid */
1951 CAPA_OPC_OSS_WRITE = 1<<5, /* write oss object data */
1952 CAPA_OPC_OSS_READ = 1<<6, /* read oss object data */
1953 CAPA_OPC_OSS_TRUNC = 1<<7, /* truncate oss object */
1954 CAPA_OPC_META_WRITE = 1<<8, /* write object meta data */
1955 CAPA_OPC_META_READ = 1<<9, /* read object meta data */
1959 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
1960 #define CAPA_OPC_MDS_ONLY \
1961 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
1962 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
1963 #define CAPA_OPC_OSS_ONLY \
1964 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC)
1965 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
1966 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
1968 /* MDS capability covers object capability for operations of body r/w
1969 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
1970 * while OSS capability only covers object capability for operations of
1971 * oss data(file content) r/w/truncate.
1973 static inline int capa_for_mds(struct lustre_capa *c)
1975 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
1978 static inline int capa_for_oss(struct lustre_capa *c)
1980 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
1983 /* lustre_capa.lc_flags */
1985 CAPA_FL_SHORT_EXPIRY = 1, /* short capa expiry */
1988 /* lustre_capa.lc_hmac_alg */
1990 CAPA_HMAC_ALG_SHA1 = 1, /* sha1 algorithm */
1994 #define CAPA_FL_MASK 0x00ffffff
1995 #define CAPA_HMAC_ALG_MASK 0xff000000
1997 struct lustre_capa_key {
1998 __u64 lk_mdsid; /* mds# */
1999 __u32 lk_keyid; /* key# */
2001 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /* key */
2002 } __attribute__((packed));
2004 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2006 typedef int (* renew_capa_cb_t)(struct obd_capa *, struct lustre_capa *);