1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre/lustre_idl.h
38 * Lustre wire protocol definitions.
41 * We assume all nodes are either little-endian or big-endian, and we
42 * always send messages in the sender's native format. The receiver
43 * detects the message format by checking the 'magic' field of the message
44 * (see lustre_msg_swabbed() below).
45 * ALL structs passing over the wire should be declared here. Structs
46 * that are used in interfaces with userspace should go in lustre_user.h.
48 * All structs being declared here should be built from simple fixed-size
49 * types (__u8, __u16, __u32, __u64) or be built from other types or
50 * structs also declared in this file. Similarly, all flags and magic
51 * values in those structs should also be declared here. This ensures
52 * that the Lustre wire protocol is not influenced by external dependencies.
54 * The only other acceptable items in this file are VERY SIMPLE accessor
55 * functions to avoid callers grubbing inside the structures, and the
56 * prototypes of the swabber functions for each struct. Nothing that
57 * depends on external functions or definitions should be in here.
59 * Structs must be properly aligned to put 64-bit values on an 8-byte
60 * boundary. Any structs being added here must also be added to
61 * utils/wirecheck.c and "make newwiretest" run to regenerate the
62 * utils/wiretest.c sources. This allows us to verify that wire structs
63 * have the proper alignment/size on all architectures.
65 * DO NOT CHANGE any of the structs, flags, values declared here and used
66 * in released Lustre versions. Some structs may have padding fields that
67 * can be used. Some structs might allow addition at the end (verify this
68 * in the code to ensure that new/old clients that see this larger struct
69 * do not fail, otherwise you need to implement protocol compatibility).
71 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
72 * implemented either here, inline (trivial implementations) or in
73 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
74 * endian, in-place in the message buffer.
76 * A swabber takes a single pointer argument. The caller must already have
77 * verified that the length of the message buffer >= sizeof (type).
79 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
80 * may be defined that swabs just the variable part, after the caller has
81 * verified that the message buffer is large enough.
84 #ifndef _LUSTRE_IDL_H_
85 #define _LUSTRE_IDL_H_
87 #include <libcfs/libcfs.h>
89 /* Defn's shared with user-space. */
90 #include <lustre/lustre_user.h>
91 #include <lustre/ll_fiemap.h>
96 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
97 * FOO_REPLY_PORTAL is for incoming replies on the FOO
98 * FOO_BULK_PORTAL is for incoming bulk on the FOO
101 #define CONNMGR_REQUEST_PORTAL 1
102 #define CONNMGR_REPLY_PORTAL 2
103 //#define OSC_REQUEST_PORTAL 3
104 #define OSC_REPLY_PORTAL 4
105 //#define OSC_BULK_PORTAL 5
106 #define OST_IO_PORTAL 6
107 #define OST_CREATE_PORTAL 7
108 #define OST_BULK_PORTAL 8
109 //#define MDC_REQUEST_PORTAL 9
110 #define MDC_REPLY_PORTAL 10
111 //#define MDC_BULK_PORTAL 11
112 #define MDS_REQUEST_PORTAL 12
113 //#define MDS_REPLY_PORTAL 13
114 #define MDS_BULK_PORTAL 14
115 #define LDLM_CB_REQUEST_PORTAL 15
116 #define LDLM_CB_REPLY_PORTAL 16
117 #define LDLM_CANCEL_REQUEST_PORTAL 17
118 #define LDLM_CANCEL_REPLY_PORTAL 18
119 //#define PTLBD_REQUEST_PORTAL 19
120 //#define PTLBD_REPLY_PORTAL 20
121 //#define PTLBD_BULK_PORTAL 21
122 #define MDS_SETATTR_PORTAL 22
123 #define MDS_READPAGE_PORTAL 23
124 #define MDS_MDS_PORTAL 24
126 #define MGC_REPLY_PORTAL 25
127 #define MGS_REQUEST_PORTAL 26
128 #define MGS_REPLY_PORTAL 27
129 #define OST_REQUEST_PORTAL 28
130 #define FLD_REQUEST_PORTAL 29
131 #define SEQ_METADATA_PORTAL 30
132 #define SEQ_DATA_PORTAL 31
133 #define SEQ_CONTROLLER_PORTAL 32
138 #define SVC_RUNNING 8
139 #define SVC_STOPPING 16
140 #define SVC_STOPPED 32
143 #define PTL_RPC_MSG_REQUEST 4711
144 #define PTL_RPC_MSG_ERR 4712
145 #define PTL_RPC_MSG_REPLY 4713
147 /* DON'T use swabbed values of MAGIC as magic! */
148 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
149 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
151 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
152 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
154 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
156 #define PTLRPC_MSG_VERSION 0x00000003
157 #define LUSTRE_VERSION_MASK 0xffff0000
158 #define LUSTRE_OBD_VERSION 0x00010000
159 #define LUSTRE_MDS_VERSION 0x00020000
160 #define LUSTRE_OST_VERSION 0x00030000
161 #define LUSTRE_DLM_VERSION 0x00040000
162 #define LUSTRE_LOG_VERSION 0x00050000
163 #define LUSTRE_MGS_VERSION 0x00060000
165 typedef __u64 mdsno_t;
166 typedef __u64 seqno_t;
173 static inline __u64 range_space(struct lu_range *r)
175 return r->lr_end - r->lr_start;
178 static inline void range_zero(struct lu_range *r)
180 r->lr_start = r->lr_end = 0;
183 static inline int range_within(struct lu_range *r,
186 return s >= r->lr_start && s < r->lr_end;
189 static inline void range_alloc(struct lu_range *r,
193 r->lr_start = s->lr_start;
194 r->lr_end = s->lr_start + w;
198 static inline int range_is_sane(struct lu_range *r)
200 return (r->lr_end >= r->lr_start);
203 static inline int range_is_zero(struct lu_range *r)
205 return (r->lr_start == 0 && r->lr_end == 0);
208 static inline int range_is_exhausted(struct lu_range *r)
210 return range_space(r) == 0;
213 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x]"
215 #define PRANGE(range) \
220 __u64 f_seq; /* holds fid sequence. Lustre should support 2 ^ 64
221 * objects, thus even if one sequence has one object we
222 * will never reach this value. */
223 __u32 f_oid; /* fid number within its sequence. */
224 __u32 f_ver; /* holds fid version. */
231 /* initial fid id value */
232 LUSTRE_FID_INIT_OID = 1UL
235 /* get object sequence */
236 static inline __u64 fid_seq(const struct lu_fid *fid)
242 static inline __u32 fid_oid(const struct lu_fid *fid)
247 /* get object version */
248 static inline __u32 fid_ver(const struct lu_fid *fid)
253 static inline void fid_zero(struct lu_fid *fid)
255 memset(fid, 0, sizeof(*fid));
258 /* Normal FID sequence starts from this value, i.e. 1<<33 */
259 #define FID_SEQ_START 0x200000000ULL
261 /* IDIF sequence starts from this value, i.e. 1<<32 */
262 #define IDIF_SEQ_START 0x100000000ULL
265 * Check if a fid is igif or not.
266 * \param fid the fid to be tested.
267 * \return true if the fid is a igif; otherwise false.
269 static inline int fid_is_igif(const struct lu_fid *fid)
271 return fid_seq(fid) > 0 && fid_seq(fid) < IDIF_SEQ_START;
275 * Check if a fid is idif or not.
276 * \param fid the fid to be tested.
277 * \return true if the fid is a idif; otherwise false.
279 static inline int fid_is_idif(const struct lu_fid *fid)
281 return fid_seq(fid) >= IDIF_SEQ_START && fid_seq(fid) < FID_SEQ_START;
285 * Get inode number from a igif.
286 * \param fid a igif to get inode number from.
287 * \return inode number for the igif.
289 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
295 * Get inode generation from a igif.
296 * \param fid a igif to get inode generation from.
297 * \return inode generation for the igif.
299 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
304 #define DFID "[0x%16.16"LPF64"x/0x%8.8x:0x%8.8x]"
311 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
313 /* check that all fields are converted */
314 CLASSERT(sizeof *src ==
315 sizeof fid_seq(src) +
316 sizeof fid_oid(src) + sizeof fid_ver(src));
317 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
318 dst->f_seq = cpu_to_le64(fid_seq(src));
319 dst->f_oid = cpu_to_le32(fid_oid(src));
320 dst->f_ver = cpu_to_le32(fid_ver(src));
323 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
325 /* check that all fields are converted */
326 CLASSERT(sizeof *src ==
327 sizeof fid_seq(src) +
328 sizeof fid_oid(src) + sizeof fid_ver(src));
329 dst->f_seq = le64_to_cpu(fid_seq(src));
330 dst->f_oid = le32_to_cpu(fid_oid(src));
331 dst->f_ver = le32_to_cpu(fid_ver(src));
332 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
335 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
337 /* check that all fields are converted */
338 CLASSERT(sizeof *src ==
339 sizeof fid_seq(src) +
340 sizeof fid_oid(src) + sizeof fid_ver(src));
341 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
342 dst->f_seq = cpu_to_be64(fid_seq(src));
343 dst->f_oid = cpu_to_be32(fid_oid(src));
344 dst->f_ver = cpu_to_be32(fid_ver(src));
347 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
349 /* check that all fields are converted */
350 CLASSERT(sizeof *src ==
351 sizeof fid_seq(src) +
352 sizeof fid_oid(src) + sizeof fid_ver(src));
353 dst->f_seq = be64_to_cpu(fid_seq(src));
354 dst->f_oid = be32_to_cpu(fid_oid(src));
355 dst->f_ver = be32_to_cpu(fid_ver(src));
356 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
361 * Storage representation for fids.
363 * Variable size, first byte contains the length of the whole record.
368 char fp_area[sizeof(struct lu_fid)];
371 void fid_pack(struct lu_fid_pack *pack, const struct lu_fid *fid,
372 struct lu_fid *befider);
373 int fid_unpack(const struct lu_fid_pack *pack, struct lu_fid *fid);
378 static inline int fid_is_sane(const struct lu_fid *fid)
382 ((fid_seq(fid) >= FID_SEQ_START && fid_oid(fid) != 0
383 && fid_ver(fid) == 0) ||
387 static inline int fid_is_zero(const struct lu_fid *fid)
389 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
392 extern void lustre_swab_lu_fid(struct lu_fid *fid);
393 extern void lustre_swab_lu_range(struct lu_range *range);
395 static inline int lu_fid_eq(const struct lu_fid *f0,
396 const struct lu_fid *f1)
398 /* Check that there is no alignment padding. */
399 CLASSERT(sizeof *f0 ==
400 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
401 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
402 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
403 return memcmp(f0, f1, sizeof *f0) == 0;
407 * Layout of readdir pages, as transmitted on wire.
410 struct lu_fid lde_fid;
419 __u64 ldp_hash_start;
423 struct lu_dirent ldp_entries[0];
426 enum lu_dirpage_flags {
430 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
432 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
435 return dp->ldp_entries;
438 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
440 struct lu_dirent *next;
442 if (le16_to_cpu(ent->lde_reclen) != 0)
443 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
450 static inline int lu_dirent_size(struct lu_dirent *ent)
452 if (le16_to_cpu(ent->lde_reclen) == 0) {
453 return (sizeof(*ent) +
454 le16_to_cpu(ent->lde_namelen) + 7) & ~7;
456 return le16_to_cpu(ent->lde_reclen);
459 #define DIR_END_OFF 0xfffffffffffffffeULL
461 struct lustre_handle {
464 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
466 static inline int lustre_handle_is_used(struct lustre_handle *lh)
468 return lh->cookie != 0ull;
471 static inline int lustre_handle_equal(struct lustre_handle *lh1,
472 struct lustre_handle *lh2)
474 return lh1->cookie == lh2->cookie;
477 static inline void lustre_handle_copy(struct lustre_handle *tgt,
478 struct lustre_handle *src)
480 tgt->cookie = src->cookie;
483 /* flags for lm_flags */
484 #define MSGHDR_AT_SUPPORT 0x1
486 #define lustre_msg lustre_msg_v2
487 /* we depend on this structure to be 8-byte aligned */
488 /* this type is only endian-adjusted in lustre_unpack_msg() */
489 struct lustre_msg_v2 {
501 /* without gss, ptlrpc_body is put at the first buffer. */
502 #define PTLRPC_NUM_VERSIONS 4
504 struct lustre_handle pb_handle;
511 __u64 pb_last_committed;
516 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
517 __u32 pb_service_time; /* for rep, actual service time */
520 /* VBR: pre-versions */
521 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
522 /* padding for future needs */
526 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
528 /* message body offset for lustre_msg_v2 */
529 /* ptlrpc body offset in all request/reply messages */
530 #define MSG_PTLRPC_BODY_OFF 0
532 /* normal request/reply message record offset */
533 #define REQ_REC_OFF 1
534 #define REPLY_REC_OFF 1
536 /* ldlm request message body offset */
537 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
538 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
540 /* ldlm intent lock message body offset */
541 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
542 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
544 /* ldlm reply message body offset */
545 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
546 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
548 /* Flags that are operation-specific go in the top 16 bits. */
549 #define MSG_OP_FLAG_MASK 0xffff0000
550 #define MSG_OP_FLAG_SHIFT 16
552 /* Flags that apply to all requests are in the bottom 16 bits */
553 #define MSG_GEN_FLAG_MASK 0x0000ffff
554 #define MSG_LAST_REPLAY 0x0001
555 #define MSG_RESENT 0x0002
556 #define MSG_REPLAY 0x0004
557 /* #define MSG_AT_SUPPORT 0x0008
558 * This was used in early prototypes of adaptive timeouts, and while there
559 * shouldn't be any users of that code there also isn't a need for using this
560 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
561 #define MSG_DELAY_REPLAY 0x0010
562 #define MSG_VERSION_REPLAY 0x0020
563 #define MSG_REQ_REPLAY_DONE 0x0040
564 #define MSG_LOCK_REPLAY_DONE 0x0080
567 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
570 #define MSG_CONNECT_RECOVERING 0x1
571 #define MSG_CONNECT_RECONNECT 0x2
572 #define MSG_CONNECT_REPLAYABLE 0x4
573 //#define MSG_CONNECT_PEER 0x8
574 #define MSG_CONNECT_LIBCLIENT 0x10
575 #define MSG_CONNECT_INITIAL 0x20
576 #define MSG_CONNECT_ASYNC 0x40
577 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
578 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
581 #define OBD_CONNECT_RDONLY 0x00000001ULL /* client allowed read-only access */
582 #define OBD_CONNECT_INDEX 0x00000002ULL /* connect to specific LOV idx */
583 #define OBD_CONNECT_GRANT 0x00000008ULL /* OSC acquires grant at connect */
584 #define OBD_CONNECT_SRVLOCK 0x00000010ULL /* server takes locks for client */
585 #define OBD_CONNECT_VERSION 0x00000020ULL /* Server supports versions in ocd */
586 #define OBD_CONNECT_REQPORTAL 0x00000040ULL /* Separate portal for non-IO reqs */
587 #define OBD_CONNECT_ACL 0x00000080ULL /* client uses access control lists */
588 #define OBD_CONNECT_XATTR 0x00000100ULL /* client using extended attributes*/
589 #define OBD_CONNECT_TRUNCLOCK 0x00000400ULL /* locks on server for punch b=9528 */
590 #define OBD_CONNECT_IBITS 0x00001000ULL /* support for inodebits locks */
591 #define OBD_CONNECT_JOIN 0x00002000ULL /* files can be concatenated */
592 #define OBD_CONNECT_ATTRFID 0x00004000ULL /* Server supports GetAttr By Fid */
593 #define OBD_CONNECT_NODEVOH 0x00008000ULL /* No open handle for special nodes */
594 #define OBD_CONNECT_LCL_CLIENT 0x00010000ULL /* local 1.8 client */
595 #define OBD_CONNECT_RMT_CLIENT 0x00020000ULL /* Remote 1.8 client */
596 #define OBD_CONNECT_BRW_SIZE 0x00040000ULL /* Max bytes per rpc */
597 #define OBD_CONNECT_QUOTA64 0x00080000ULL /* 64bit qunit_data.qd_count b=10707*/
598 #define OBD_CONNECT_MDS_CAPA 0x00100000ULL /* MDS capability */
599 #define OBD_CONNECT_OSS_CAPA 0x00200000ULL /* OSS capability */
600 #define OBD_CONNECT_CANCELSET 0x00400000ULL /* Early batched cancels. */
601 #define OBD_CONNECT_SOM 0x00800000ULL /* SOM feature */
602 #define OBD_CONNECT_AT 0x01000000ULL /* client uses adaptive timeouts */
603 #define OBD_CONNECT_LRU_RESIZE 0x02000000ULL /* Lru resize feature. */
604 #define OBD_CONNECT_MDS_MDS 0x04000000ULL /* MDS-MDS connection*/
605 #define OBD_CONNECT_REAL 0x08000000ULL /* real connection */
606 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*shrink/enlarge qunit size
608 #define OBD_CONNECT_CKSUM 0x20000000ULL /* support several cksum algos */
609 #define OBD_CONNECT_FID 0x40000000ULL /* FID is supported by server */
611 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
612 * and lustre/utils/wirecheck.c */
614 #ifdef HAVE_LRU_RESIZE_SUPPORT
615 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
617 #define LRU_RESIZE_CONNECT_FLAG 0
620 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
621 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
622 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
623 OBD_CONNECT_NODEVOH |/* OBD_CONNECT_ATTRFID |*/\
624 OBD_CONNECT_LCL_CLIENT | \
625 OBD_CONNECT_RMT_CLIENT | \
626 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | \
627 OBD_CONNECT_MDS_MDS | OBD_CONNECT_CANCELSET | \
629 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_AT)
630 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
631 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
632 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
633 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
634 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET | \
635 OBD_CONNECT_CKSUM | LRU_RESIZE_CONNECT_FLAG | \
637 #define ECHO_CONNECT_SUPPORTED (0)
638 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT)
640 #define MAX_QUOTA_COUNT32 (0xffffffffULL)
642 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
643 ((patch)<<8) + (fix))
644 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
645 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
646 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
647 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
649 /* This structure is used for both request and reply.
651 * If we eventually have separate connect data for different types, which we
652 * almost certainly will, then perhaps we stick a union in here. */
653 struct obd_connect_data {
654 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
655 __u32 ocd_version; /* lustre release version number */
656 __u32 ocd_grant; /* initial cache grant amount (bytes) */
657 __u32 ocd_index; /* LOV index to connect to */
658 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
659 __u64 ocd_ibits_known; /* inode bits this client understands */
660 __u32 ocd_nllu; /* non-local-lustre-user */
661 __u32 ocd_nllg; /* non-local-lustre-group */
662 __u64 ocd_transno; /* first transno from client to be replayed */
663 __u32 ocd_group; /* MDS group on OST */
664 __u32 ocd_cksum_types; /* supported checksum algorithms */
665 __u64 padding1; /* also fix lustre_swab_connect */
666 __u64 padding2; /* also fix lustre_swab_connect */
669 extern void lustre_swab_connect(struct obd_connect_data *ocd);
672 * Supported checksum algorithms. Up to 32 checksum types are supported.
673 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
674 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
675 * algorithm and also the OBD_FL_CKSUM* flags.
678 OBD_CKSUM_CRC32 = 0x00000001,
679 OBD_CKSUM_ADLER = 0x00000002,
683 * OST requests: OBDO & OBD request records
688 OST_REPLY = 0, /* reply ? */
708 #define OST_FIRST_OPC OST_REPLY
710 typedef __u64 obd_id;
711 typedef __u64 obd_gr;
712 typedef __u64 obd_time;
713 typedef __u64 obd_size;
714 typedef __u64 obd_off;
715 typedef __u64 obd_blocks;
716 typedef __u32 obd_blksize;
717 typedef __u32 obd_mode;
718 typedef __u32 obd_uid;
719 typedef __u32 obd_gid;
720 typedef __u32 obd_flag;
721 typedef __u64 obd_valid;
722 typedef __u32 obd_count;
724 #define OBD_FL_INLINEDATA (0x00000001)
725 #define OBD_FL_OBDMDEXISTS (0x00000002)
726 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
727 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
728 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
729 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
730 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
731 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
732 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
735 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
736 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
739 #define OBD_FL_TRUNCLOCK (0x00000800)
744 #define OBD_FL_CKSUM_CRC32 (0x00001000)
745 #define OBD_FL_CKSUM_ADLER (0x00002000)
746 #define OBD_FL_CKSUM_ALL (OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER)
749 * This should not be smaller than sizeof(struct lustre_handle) + sizeof(struct
750 * llog_cookie) + sizeof(struct ll_fid). Nevertheless struct ll_fid is not
751 * longer stored in o_inline, we keep this just for case.
753 #define OBD_INLINESZ 80
755 /* Note: 64-bit types are 64-bit aligned in structure */
757 obd_valid o_valid; /* hot fields in this obdo */
761 obd_size o_size; /* o_size-o_blocks == ost_lvb */
765 obd_blocks o_blocks; /* brw: cli sent cached bytes */
768 /* 32-bit fields start here: keep an even number of them via padding */
769 obd_blksize o_blksize; /* optimal IO blocksize */
770 obd_mode o_mode; /* brw: cli sent cache remain */
774 obd_count o_nlink; /* brw: checksum */
775 obd_count o_generation;
776 obd_count o_misc; /* brw: o_dropped */
777 __u32 o_easize; /* epoch in ost writes */
779 __u32 o_stripe_idx; /* holds stripe idx */
781 char o_inline[OBD_INLINESZ];
782 /* lustre_handle + llog_cookie */
785 #define o_dirty o_blocks
786 #define o_undirty o_mode
787 #define o_dropped o_misc
788 #define o_cksum o_nlink
790 extern void lustre_swab_obdo (struct obdo *o);
793 #define LOV_MAGIC_V1 0x0BD10BD0
794 #define LOV_MAGIC LOV_MAGIC_V1
795 #define LOV_MAGIC_JOIN 0x0BD20BD0
797 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
798 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
799 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
800 #define LOV_PATTERN_CMOBD 0x200
802 #define LOV_OBJECT_GROUP_DEFAULT ~0ULL
803 #define LOV_OBJECT_GROUP_CLEAR 0ULL
805 #define lov_ost_data lov_ost_data_v1
806 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
807 __u64 l_object_id; /* OST object ID */
808 __u64 l_object_gr; /* OST object group (creating MDS number) */
809 __u32 l_ost_gen; /* generation of this l_ost_idx */
810 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
813 #define lov_mds_md lov_mds_md_v1
814 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
815 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
816 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
817 __u64 lmm_object_id; /* LOV object ID */
818 __u64 lmm_object_gr; /* LOV object group */
819 __u32 lmm_stripe_size; /* size of stripe in bytes */
820 __u32 lmm_stripe_count; /* num stripes in use for this object */
821 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
824 extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm);
826 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
827 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
829 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
830 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
831 #define XATTR_NAME_LOV "trusted.lov"
833 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
834 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
835 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
836 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
837 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
838 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
839 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
840 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
841 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
842 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
843 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
844 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
845 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
846 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
847 #define OBD_MD_FLINLINE (0x00008000ULL) /* inline data */
848 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
849 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
850 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
851 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file handle */
852 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
853 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
854 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
855 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
856 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
857 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
858 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
859 /* ->mds if epoch opens or closes */
860 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
861 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
862 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
863 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
864 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
866 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
867 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
868 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
870 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
871 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
872 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
873 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
874 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
875 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
876 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
877 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
878 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
880 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
881 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
882 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
883 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
885 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
886 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
887 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
888 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
889 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
891 static inline struct lustre_handle *obdo_handle(struct obdo *oa)
893 return (struct lustre_handle *)oa->o_inline;
896 static inline struct llog_cookie *obdo_logcookie(struct obdo *oa)
898 return (struct llog_cookie *)(oa->o_inline +
899 sizeof(struct lustre_handle));
901 /* don't forget obdo_fid which is way down at the bottom so it can
902 * come after the definition of llog_cookie */
915 __u32 os_state; /* positive error code on server */
927 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
928 #define OBD_STATFS_NODELAY 0x0001 /* requests should be send without delay
929 * and resends for avoid deadlocks */
931 #define OBD_STATFS_FROM_CACHE 0x0002 /* the statfs callback should not update
934 /* ost_body.data values for OST_BRW */
936 #define OBD_BRW_READ 0x01
937 #define OBD_BRW_WRITE 0x02
938 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
939 #define OBD_BRW_SYNC 0x08
940 #define OBD_BRW_CHECK 0x10
941 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
942 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
943 #define OBD_BRW_DROP 0x80 /* drop the page after IO */
944 #define OBD_BRW_NOQUOTA 0x100
945 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
947 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
949 #define OST_MIN_PRECREATE 32
950 #define OST_MAX_PRECREATE 20000
959 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
961 /* multiple of 8 bytes => can array */
962 struct niobuf_remote {
968 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
970 /* request structure for OST's */
976 /* Key for FIEMAP to be used in get_info calls */
977 struct ll_fiemap_info_key {
980 struct ll_user_fiemap fiemap;
983 extern void lustre_swab_ost_body (struct ost_body *b);
984 extern void lustre_swab_ost_last_id(obd_id *id);
985 extern void lustre_swab_fiemap(struct ll_user_fiemap *fiemap);
987 /* lock value block communicated between the filter and llite */
989 /* OST_LVB_ERR_INIT is needed because the return code in rc is
990 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
991 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
992 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
993 #define OST_LVB_IS_ERR(blocks) \
994 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
995 #define OST_LVB_SET_ERR(blocks, rc) \
996 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
997 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1007 extern void lustre_swab_ost_lvb(struct ost_lvb *);
1016 MDS_GETATTR_NAME = 34,
1021 MDS_DISCONNECT = 39,
1027 MDS_DONE_WRITING = 45,
1029 MDS_QUOTACHECK = 47,
1032 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1038 #define MDS_FIRST_OPC MDS_GETATTR
1055 } mds_reint_t, mdt_reint_t;
1057 extern void lustre_swab_generic_32s (__u32 *val);
1059 /* the disposition of the intent outlines what was executed */
1060 #define DISP_IT_EXECD 0x00000001
1061 #define DISP_LOOKUP_EXECD 0x00000002
1062 #define DISP_LOOKUP_NEG 0x00000004
1063 #define DISP_LOOKUP_POS 0x00000008
1064 #define DISP_OPEN_CREATE 0x00000010
1065 #define DISP_OPEN_OPEN 0x00000020
1066 #define DISP_ENQ_COMPLETE 0x00400000
1067 #define DISP_ENQ_OPEN_REF 0x00800000
1068 #define DISP_ENQ_CREATE_REF 0x01000000
1069 #define DISP_OPEN_LOCK 0x02000000
1071 /* INODE LOCK PARTS */
1072 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1073 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1074 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1076 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
1077 #define MDS_INODELOCK_MAXSHIFT 2
1078 /* This FULL lock is useful to take on unlink sort of operations */
1079 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1082 __u64 id; /* holds object id */
1083 __u32 generation; /* holds object generation */
1085 __u32 f_type; /* holds object type or stripe idx when passing it to
1086 * OST for saving into EA. */
1089 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1091 #define MDS_STATUS_CONN 1
1092 #define MDS_STATUS_LOV 2
1094 struct mds_status_req {
1099 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
1101 /* mdt_thread_info.mti_flags. */
1103 /* The flag indicates Size-on-MDS attributes are changed. */
1104 MF_SOM_CHANGE = (1 << 0),
1105 /* Flags indicates an epoch opens or closes. */
1106 MF_EPOCH_OPEN = (1 << 1),
1107 MF_EPOCH_CLOSE = (1 << 2),
1108 MF_MDC_CANCEL_FID1 = (1 << 3),
1109 MF_MDC_CANCEL_FID2 = (1 << 4),
1110 MF_MDC_CANCEL_FID3 = (1 << 5),
1111 MF_MDC_CANCEL_FID4 = (1 << 6),
1114 #define MF_SOM_LOCAL_FLAGS (MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID2 | \
1115 MF_MDC_CANCEL_FID3 | MF_MDC_CANCEL_FID4)
1117 #define MDS_BFLAG_UNCOMMITTED_WRITES 0x1
1118 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
1120 /* these should be identical to their EXT3_*_FL counterparts, and are
1121 * redefined here only to avoid dragging in ext3_fs.h */
1122 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
1123 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
1124 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
1125 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
1126 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1129 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
1130 * and we need to decode these into local S_* flags in the inode. Otherwise
1131 * we pass flags straight through (see bug 9486). */
1132 static inline int ll_ext_to_inode_flags(int flags)
1134 return (flags & MDS_BFLAG_EXT_FLAGS) ?
1135 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
1136 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
1137 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
1138 #if defined(S_DIRSYNC)
1139 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1141 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1142 (flags & ~MDS_BFLAG_EXT_FLAGS);
1145 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1146 * and we pass these straight through. Otherwise we need to convert from
1147 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1148 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1150 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1151 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1152 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1153 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1154 #if defined(S_DIRSYNC)
1155 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1157 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1164 struct lustre_handle handle;
1166 __u64 size; /* Offset, in the case of MDS_READPAGE */
1170 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1172 __u64 ino; /* for 1.6 compatibility */
1179 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1181 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1182 __u32 generation; /* for 1.6 compatibility */
1187 __u32 max_cookiesize;
1188 __u32 padding_4; /* also fix lustre_swab_mdt_body */
1194 struct lustre_handle handle;
1196 __u64 size; /* Offset, in the case of MDS_READPAGE */
1200 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1209 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1211 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1217 __u32 max_cookiesize;
1218 __u32 padding_4; /* also fix lustre_swab_mds_body */
1221 extern void lustre_swab_mds_body (struct mds_body *b);
1222 extern void lustre_swab_mdt_body (struct mdt_body *b);
1225 struct lustre_handle handle;
1231 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1233 #define Q_QUOTACHECK 0x800100
1234 #define Q_INITQUOTA 0x800101 /* init slave limits */
1235 #define Q_GETOINFO 0x800102 /* get obd quota info */
1236 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1238 #define Q_TYPESET(oqc, type) \
1239 ((oqc)->qc_type == type || (oqc)->qc_type == UGQUOTA)
1241 #define Q_GETOCMD(oqc) \
1242 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1244 struct obd_quotactl {
1249 struct obd_dqinfo qc_dqinfo;
1250 struct obd_dqblk qc_dqblk;
1253 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1255 /* inode access permission for remote user, the inode info are omitted,
1256 * for client knows them. */
1257 struct mds_remote_perm {
1262 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1265 /* permissions for md_perm.mp_perm */
1267 CFS_SETUID_PERM = 0x01,
1268 CFS_SETGID_PERM = 0x02,
1269 CFS_SETGRP_PERM = 0x04,
1270 CFS_RMTACL_PERM = 0x08
1273 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1275 struct mdt_remote_perm {
1280 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1283 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1285 struct mds_rec_setattr {
1292 struct ll_fid sa_fid;
1293 __u64 sa_valid; /* MDS_ATTR_* attributes */
1300 __u32 sa_attr_flags;
1301 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1304 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1306 struct mdt_rec_setattr {
1313 struct lu_fid sa_fid;
1322 __u32 sa_attr_flags;
1330 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1333 * Attribute flags used in mds_rec_setattr::sa_valid.
1334 * The kernel's #defines for ATTR_* should not be used over the network
1335 * since the client and MDS may run different kernels (see bug 13828)
1336 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1338 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1339 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1340 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1341 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1342 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1343 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1344 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1345 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1346 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1347 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1348 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1349 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1350 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1351 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1352 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1353 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1356 #define FMODE_READ 00000001
1357 #define FMODE_WRITE 00000002
1360 #define FMODE_EPOCH 01000000
1361 #define FMODE_EPOCHLCK 02000000
1362 #define FMODE_SOM 04000000
1363 #define FMODE_CLOSED 0
1365 #define MDS_OPEN_CREATED 00000010
1366 #define MDS_OPEN_CROSS 00000020
1368 #define MDS_FMODE_EXEC 00000004
1369 #define MDS_OPEN_CREAT 00000100
1370 #define MDS_OPEN_EXCL 00000200
1371 #define MDS_OPEN_TRUNC 00001000
1372 #define MDS_OPEN_APPEND 00002000
1373 #define MDS_OPEN_SYNC 00010000
1374 #define MDS_OPEN_DIRECTORY 00200000
1376 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1377 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1378 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1379 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1380 * with default ACL */
1381 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1382 * actually, this is for create, not
1383 * conflict with other open flags */
1384 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1385 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1386 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1388 /* permission for create non-directory file */
1389 #define MAY_CREATE (1 << 7)
1390 /* permission for create directory file */
1391 #define MAY_LINK (1 << 8)
1392 /* permission for delete from the directory */
1393 #define MAY_UNLINK (1 << 9)
1394 /* source's permission for rename */
1395 #define MAY_RENAME_SRC (1 << 10)
1396 /* target's permission for rename */
1397 #define MAY_RENAME_TAR (1 << 11)
1398 /* part (parent's) VTX permission check */
1399 #define MAY_VTX_PART (1 << 12)
1400 /* full VTX permission check */
1401 #define MAY_VTX_FULL (1 << 13)
1402 /* lfs rgetfacl permission check */
1403 #define MAY_RGETFACL (1 << 14)
1406 MDS_CHECK_SPLIT = 1 << 0,
1407 MDS_CROSS_REF = 1 << 1,
1408 MDS_VTX_BYPASS = 1 << 2,
1409 MDS_PERM_BYPASS = 1 << 3,
1413 struct mds_rec_join {
1414 struct ll_fid jr_fid;
1418 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1420 struct mdt_rec_join {
1421 struct lu_fid jr_fid;
1425 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1427 struct mds_rec_create {
1432 __u32 cr_flags; /* for use with open */
1434 struct ll_fid cr_fid;
1435 struct ll_fid cr_replayfid;
1439 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1440 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1441 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1442 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1443 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1446 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1448 struct mdt_rec_create {
1455 struct lu_fid cr_fid1;
1456 struct lu_fid cr_fid2;
1457 struct lustre_handle cr_old_handle; /* u64 handle in case of open replay */
1461 __u64 cr_padding_1; /* pad for 64 bits*/
1464 __u32 cr_flags; /* for use with open */
1470 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1472 struct mds_rec_link {
1479 struct ll_fid lk_fid1;
1480 struct ll_fid lk_fid2;
1482 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1483 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1484 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1485 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1488 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1490 struct mdt_rec_link {
1497 struct lu_fid lk_fid1;
1498 struct lu_fid lk_fid2;
1512 struct mds_rec_unlink {
1519 struct ll_fid ul_fid1;
1520 struct ll_fid ul_fid2;
1522 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1523 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1524 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1525 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1528 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1530 struct mdt_rec_unlink {
1537 struct lu_fid ul_fid1;
1538 struct lu_fid ul_fid2;
1552 struct mds_rec_rename {
1559 struct ll_fid rn_fid1;
1560 struct ll_fid rn_fid2;
1562 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1563 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1564 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1565 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1568 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1570 struct mdt_rec_rename {
1577 struct lu_fid rn_fid1;
1578 struct lu_fid rn_fid2;
1584 __u32 rn_bias; /* some operation flags */
1585 __u32 rn_mode; /* cross-ref rename has mode */
1592 struct mdt_rec_setxattr {
1599 struct lu_fid sx_fid;
1600 __u64 sx_padding_1; /* These three members are lu_fid size */
1612 __u32 sx_padding_10;
1613 __u32 sx_padding_11;
1616 struct mdt_rec_reint {
1623 struct lu_fid rr_fid1;
1624 struct lu_fid rr_fid2;
1632 __u32 rr_padding_1; /* also fix lustre_swab_mdt_rec_reint */
1633 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
1634 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
1635 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1638 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
1641 __u32 ld_tgt_count; /* how many MDS's */
1642 __u32 ld_active_tgt_count; /* how many active */
1643 struct obd_uuid ld_uuid;
1646 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1653 extern void lustre_swab_md_fld (struct md_fld *mf);
1658 FLD_FIRST_OPC = FLD_QUERY
1664 SEQ_FIRST_OPC = SEQ_QUERY
1668 SEQ_ALLOC_SUPER = 0,
1673 * LOV data structures
1676 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1677 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1678 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
1680 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1681 /* The size of the buffer the lov/mdc reserves for the
1682 * array of UUIDs returned by the MDS. With the current
1683 * protocol, this will limit the max number of OSTs per LOV */
1685 #define LOV_DESC_MAGIC 0xB0CCDE5C
1687 /* LOV settings descriptor (should only contain static info) */
1689 __u32 ld_tgt_count; /* how many OBD's */
1690 __u32 ld_active_tgt_count; /* how many active */
1691 __u32 ld_default_stripe_count; /* how many objects are used */
1692 __u32 ld_pattern; /* default PATTERN_RAID0 */
1693 __u64 ld_default_stripe_size; /* in bytes */
1694 __u64 ld_default_stripe_offset; /* in bytes */
1695 __u32 ld_padding_0; /* unused */
1696 __u32 ld_qos_maxage; /* in second */
1697 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1698 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1699 struct obd_uuid ld_uuid;
1702 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1704 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1709 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1714 LDLM_BL_CALLBACK = 104,
1715 LDLM_CP_CALLBACK = 105,
1716 LDLM_GL_CALLBACK = 106,
1719 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1721 #define RES_NAME_SIZE 4
1722 struct ldlm_res_id {
1723 __u64 name[RES_NAME_SIZE];
1726 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1741 #define LCK_MODE_NUM 7
1751 #define LDLM_MIN_TYPE LDLM_PLAIN
1753 struct ldlm_extent {
1759 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
1760 struct ldlm_extent *ex2)
1762 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
1765 struct ldlm_inodebits {
1772 __u64 blocking_export; /* not actually used over the wire */
1773 __u32 blocking_pid; /* not actually used over the wire */
1777 /* it's important that the fields of the ldlm_extent structure match
1778 * the first fields of the ldlm_flock structure because there is only
1779 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1780 * this ever changes we will need to swab the union differently based
1781 * on the resource type. */
1784 struct ldlm_extent l_extent;
1785 struct ldlm_flock l_flock;
1786 struct ldlm_inodebits l_inodebits;
1787 } ldlm_policy_data_t;
1789 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1791 struct ldlm_intent {
1795 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1797 struct ldlm_resource_desc {
1798 ldlm_type_t lr_type;
1799 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1800 struct ldlm_res_id lr_name;
1803 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1805 struct ldlm_lock_desc {
1806 struct ldlm_resource_desc l_resource;
1807 ldlm_mode_t l_req_mode;
1808 ldlm_mode_t l_granted_mode;
1809 ldlm_policy_data_t l_policy_data;
1812 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1814 #define LDLM_LOCKREQ_HANDLES 2
1815 #define LDLM_ENQUEUE_CANCEL_OFF 1
1817 struct ldlm_request {
1820 struct ldlm_lock_desc lock_desc;
1821 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
1824 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1826 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
1827 * Otherwise, 2 are available. */
1828 #define ldlm_request_bufsize(count,type) \
1830 int _avail = LDLM_LOCKREQ_HANDLES; \
1831 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
1832 sizeof(struct ldlm_request) + \
1833 (count > _avail ? count - _avail : 0) * \
1834 sizeof(struct lustre_handle); \
1839 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1840 struct ldlm_lock_desc lock_desc;
1841 struct lustre_handle lock_handle;
1842 __u64 lock_policy_res1;
1843 __u64 lock_policy_res2;
1846 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1849 * Opcodes for mountconf (mgs and mgc)
1854 MGS_EXCEPTION, /* node died, etc. */
1855 MGS_TARGET_REG, /* whenever target starts up */
1860 #define MGS_FIRST_OPC MGS_CONNECT
1862 #define MGS_PARAM_MAXLEN 1024
1863 #define KEY_SET_INFO "set_info"
1865 struct mgs_send_param {
1866 char mgs_param[MGS_PARAM_MAXLEN];
1869 /* We pass this info to the MGS so it can write config logs */
1870 #define MTI_NAME_MAXLEN 64
1871 #define MTI_PARAM_MAXLEN 4096
1872 #define MTI_NIDS_MAX 32
1873 struct mgs_target_info {
1874 __u32 mti_lustre_ver;
1875 __u32 mti_stripe_index;
1876 __u32 mti_config_ver;
1878 __u32 mti_nid_count;
1879 __u32 padding; /* 64 bit align */
1880 char mti_fsname[MTI_NAME_MAXLEN];
1881 char mti_svname[MTI_NAME_MAXLEN];
1882 char mti_uuid[sizeof(struct obd_uuid)];
1883 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
1884 char mti_params[MTI_PARAM_MAXLEN];
1887 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
1889 /* Config marker flags (in config log) */
1890 #define CM_START 0x01
1892 #define CM_SKIP 0x04
1893 #define CM_UPGRADE146 0x08
1894 #define CM_EXCLUDE 0x10
1895 #define CM_START_SKIP (CM_START | CM_SKIP)
1898 __u32 cm_step; /* aka config version */
1900 __u32 cm_vers; /* lustre release version number */
1901 __u32 padding; /* 64 bit align */
1902 time_t cm_createtime; /*when this record was first created */
1903 time_t cm_canceltime; /*when this record is no longer valid*/
1904 char cm_tgtname[MTI_NAME_MAXLEN];
1905 char cm_comment[MTI_NAME_MAXLEN];
1909 * Opcodes for multiple servers.
1918 #define OBD_FIRST_OPC OBD_PING
1920 /* catalog of log objects */
1922 /* Identifier for a single log object */
1927 } __attribute__((packed));
1929 /* Records written to the CATALOGS list */
1930 #define CATLIST "CATALOGS"
1932 struct llog_logid lci_logid;
1936 } __attribute__((packed));
1938 /*join file lov mds md*/
1939 struct lov_mds_md_join {
1940 struct lov_mds_md lmmj_md;
1941 /*join private info*/
1942 struct llog_logid lmmj_array_id; /*array object id*/
1943 __u32 lmmj_extent_count; /*array extent count*/
1946 /* Log data record types - there is no specific reason that these need to
1947 * be related to the RPC opcodes, but no reason not to (may be handy later?)
1949 #define LLOG_OP_MAGIC 0x10600000
1950 #define LLOG_OP_MASK 0xfff00000
1953 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
1954 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
1955 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
1956 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
1957 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
1958 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
1959 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
1960 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
1961 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
1962 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
1963 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
1967 * for now, continue to support old pad records which have 0 for their
1968 * type but still need to be swabbed for their length
1970 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
1971 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
1972 __swab32(LLOG_OP_MAGIC) || \
1973 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
1975 /* Log record header - stored in little endian order.
1976 * Each record must start with this struct, end with a llog_rec_tail,
1977 * and be a multiple of 256 bits in size.
1979 struct llog_rec_hdr {
1986 struct llog_rec_tail {
1991 struct llog_logid_rec {
1992 struct llog_rec_hdr lid_hdr;
1993 struct llog_logid lid_id;
1999 struct llog_rec_tail lid_tail;
2000 } __attribute__((packed));
2002 /* MDS extent description
2003 * It is for joined file extent info, each extent info for joined file
2004 * just like (start, end, lmm).
2006 struct mds_extent_desc {
2007 __u64 med_start; /* extent start */
2008 __u64 med_len; /* extent length */
2009 struct lov_mds_md med_lmm; /* extent's lmm */
2011 /*Joined file array extent log record*/
2012 struct llog_array_rec {
2013 struct llog_rec_hdr lmr_hdr;
2014 struct mds_extent_desc lmr_med;
2015 struct llog_rec_tail lmr_tail;
2018 struct llog_create_rec {
2019 struct llog_rec_hdr lcr_hdr;
2020 struct ll_fid lcr_fid;
2024 struct llog_rec_tail lcr_tail;
2025 } __attribute__((packed));
2027 struct llog_orphan_rec {
2028 struct llog_rec_hdr lor_hdr;
2032 struct llog_rec_tail lor_tail;
2033 } __attribute__((packed));
2035 struct llog_unlink_rec {
2036 struct llog_rec_hdr lur_hdr;
2040 struct llog_rec_tail lur_tail;
2041 } __attribute__((packed));
2043 struct llog_setattr_rec {
2044 struct llog_rec_hdr lsr_hdr;
2050 struct llog_rec_tail lsr_tail;
2051 } __attribute__((packed));
2053 struct llog_size_change_rec {
2054 struct llog_rec_hdr lsc_hdr;
2055 struct ll_fid lsc_fid;
2058 struct llog_rec_tail lsc_tail;
2059 } __attribute__((packed));
2064 } __attribute__((packed));
2066 struct llog_gen_rec {
2067 struct llog_rec_hdr lgr_hdr;
2068 struct llog_gen lgr_gen;
2069 struct llog_rec_tail lgr_tail;
2071 /* On-disk header structure of each log object, stored in little endian order */
2072 #define LLOG_CHUNK_SIZE 8192
2073 #define LLOG_HEADER_SIZE (96)
2074 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2076 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2078 /* flags for the logs */
2079 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
2080 #define LLOG_F_IS_CAT 0x2
2081 #define LLOG_F_IS_PLAIN 0x4
2083 struct llog_log_hdr {
2084 struct llog_rec_hdr llh_hdr;
2085 __u64 llh_timestamp;
2087 __u32 llh_bitmap_offset;
2091 /* for a catalog the first plain slot is next to it */
2092 struct obd_uuid llh_tgtuuid;
2093 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2094 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2095 struct llog_rec_tail llh_tail;
2096 } __attribute__((packed));
2098 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
2099 llh->llh_bitmap_offset - \
2100 sizeof(llh->llh_tail)) * 8)
2102 /* log cookies are used to reference a specific log file and a record therein */
2103 struct llog_cookie {
2104 struct llog_logid lgc_lgl;
2108 } __attribute__((packed));
2111 enum llogd_rpc_ops {
2112 LLOG_ORIGIN_HANDLE_CREATE = 501,
2113 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2114 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2115 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2116 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2117 LLOG_ORIGIN_CONNECT = 506,
2118 LLOG_CATINFO = 507, /* for lfs catinfo */
2119 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2120 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2122 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2126 struct llog_logid lgd_logid;
2128 __u32 lgd_llh_flags;
2130 __u32 lgd_saved_index;
2132 __u64 lgd_cur_offset;
2133 } __attribute__((packed));
2135 struct llogd_conn_body {
2136 struct llog_gen lgdc_gen;
2137 struct llog_logid lgdc_logid;
2138 __u32 lgdc_ctxt_idx;
2139 } __attribute__((packed));
2141 struct lov_user_ost_data_join { /* per-stripe data structure */
2142 __u64 l_extent_start; /* extent start*/
2143 __u64 l_extent_end; /* extent end*/
2144 __u64 l_object_id; /* OST object ID */
2145 __u64 l_object_gr; /* OST object group (creating MDS number) */
2146 __u32 l_ost_gen; /* generation of this OST index */
2147 __u32 l_ost_idx; /* OST index in LOV */
2148 } __attribute__((packed));
2150 struct lov_user_md_join { /* LOV EA user data (host-endian) */
2151 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
2152 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
2153 __u64 lmm_object_id; /* LOV object ID */
2154 __u64 lmm_object_gr; /* LOV object group */
2155 __u32 lmm_stripe_size; /* size of stripe in bytes */
2156 __u32 lmm_stripe_count; /* num stripes in use for this object */
2157 __u32 lmm_extent_count; /* extent count of lmm*/
2158 __u64 lmm_tree_id; /* mds tree object id */
2159 __u64 lmm_tree_gen; /* mds tree object gen */
2160 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
2161 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
2162 } __attribute__((packed));
2164 extern void lustre_swab_lov_user_md(struct lov_user_md *lum);
2165 extern void lustre_swab_lov_user_md_objects(struct lov_user_md *lum);
2166 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
2169 extern void lustre_swab_llogd_body (struct llogd_body *d);
2170 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
2171 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
2172 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
2173 struct llog_rec_tail *tail);
2176 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
2178 /* quota. fixed by tianzy for bug10707 */
2179 #define QUOTA_IS_GRP 0X1UL /* 0 is user, 1 is group. Used by qd_flags*/
2180 #define QUOTA_IS_BLOCK 0x2UL /* 0 is inode, 1 is block. Used by qd_flags*/
2183 __u32 qd_id; /* ID appiles to (uid, gid) */
2184 __u32 qd_flags; /* Quota type (USRQUOTA, GRPQUOTA) occupy one bit;
2185 * Block quota or file quota occupy one bit */
2186 __u64 qd_count; /* acquire/release count (bytes for block quota) */
2189 struct qunit_data_old {
2190 __u32 qd_id; /* ID appiles to (uid, gid) */
2191 __u32 qd_type; /* Quota type (USRQUOTA, GRPQUOTA) */
2192 __u32 qd_count; /* acquire/release count (bytes for block quota) */
2193 __u32 qd_isblk; /* Block quota or file quota */
2196 extern void lustre_swab_qdata(struct qunit_data *d);
2197 extern void lustre_swab_qdata_old(struct qunit_data_old *d);
2198 extern struct qunit_data *lustre_quota_old_to_new(struct qunit_data_old *d);
2199 extern struct qunit_data_old *lustre_quota_new_to_old(struct qunit_data *d);
2206 #define JOIN_FILE_ALIGN 4096
2208 /* security opcodes */
2211 SEC_CTX_INIT_CONT = 802,
2214 SEC_FIRST_OPC = SEC_CTX_INIT
2218 * capa related definitions
2220 #define CAPA_HMAC_MAX_LEN 64
2221 #define CAPA_HMAC_KEY_MAX_LEN 56
2223 /* NB take care when changing the sequence of elements this struct,
2224 * because the offset info is used in find_capa() */
2225 struct lustre_capa {
2226 struct lu_fid lc_fid; /* fid */
2227 __u64 lc_opc; /* operations allowed */
2228 __u32 lc_uid; /* uid, it is obsolete, but maybe used in
2229 * future, reserve it for 64-bits aligned.*/
2230 __u32 lc_flags; /* HMAC algorithm & flags */
2231 __u32 lc_keyid; /* key used for the capability */
2232 __u32 lc_timeout; /* capa timeout value (sec) */
2233 __u64 lc_expiry; /* expiry time (sec) */
2234 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /* HMAC */
2235 } __attribute__((packed));
2237 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
2239 /* lustre_capa.lc_opc */
2241 CAPA_OPC_BODY_WRITE = 1<<0, /* write object data */
2242 CAPA_OPC_BODY_READ = 1<<1, /* read object data */
2243 CAPA_OPC_INDEX_LOOKUP = 1<<2, /* lookup object fid */
2244 CAPA_OPC_INDEX_INSERT = 1<<3, /* insert object fid */
2245 CAPA_OPC_INDEX_DELETE = 1<<4, /* delete object fid */
2246 CAPA_OPC_OSS_WRITE = 1<<5, /* write oss object data */
2247 CAPA_OPC_OSS_READ = 1<<6, /* read oss object data */
2248 CAPA_OPC_OSS_TRUNC = 1<<7, /* truncate oss object */
2249 CAPA_OPC_META_WRITE = 1<<8, /* write object meta data */
2250 CAPA_OPC_META_READ = 1<<9, /* read object meta data */
2254 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2255 #define CAPA_OPC_MDS_ONLY \
2256 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2257 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2258 #define CAPA_OPC_OSS_ONLY \
2259 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC)
2260 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2261 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2263 /* MDS capability covers object capability for operations of body r/w
2264 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
2265 * while OSS capability only covers object capability for operations of
2266 * oss data(file content) r/w/truncate.
2268 static inline int capa_for_mds(struct lustre_capa *c)
2270 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2273 static inline int capa_for_oss(struct lustre_capa *c)
2275 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2278 /* lustre_capa.lc_hmac_alg */
2280 CAPA_HMAC_ALG_SHA1 = 1, /* sha1 algorithm */
2284 #define CAPA_FL_MASK 0x00ffffff
2285 #define CAPA_HMAC_ALG_MASK 0xff000000
2287 struct lustre_capa_key {
2288 __u64 lk_mdsid; /* mds# */
2289 __u32 lk_keyid; /* key# */
2291 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /* key */
2292 } __attribute__((packed));
2294 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2296 /* quota check function */
2297 #define QUOTA_RET_OK 0 /* return successfully */
2298 #define QUOTA_RET_NOQUOTA 1 /* not support quota */
2299 #define QUOTA_RET_NOLIMIT 2 /* quota limit isn't set */
2300 #define QUOTA_RET_ACQUOTA 3 /* need to acquire extra quota */