4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2015, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * Lustre wire protocol definitions.
35 /** \defgroup lustreidl lustreidl
37 * Lustre wire protocol definitions.
39 * ALL structs passing over the wire should be declared here. Structs
40 * that are used in interfaces with userspace should go in lustre_user.h.
42 * All structs being declared here should be built from simple fixed-size
43 * types (__u8, __u16, __u32, __u64) or be built from other types or
44 * structs also declared in this file. Similarly, all flags and magic
45 * values in those structs should also be declared here. This ensures
46 * that the Lustre wire protocol is not influenced by external dependencies.
48 * The only other acceptable items in this file are VERY SIMPLE accessor
49 * functions to avoid callers grubbing inside the structures. Nothing that
50 * depends on external functions or definitions should be in here.
52 * Structs must be properly aligned to put 64-bit values on an 8-byte
53 * boundary. Any structs being added here must also be added to
54 * utils/wirecheck.c and "make newwiretest" run to regenerate the
55 * utils/wiretest.c sources. This allows us to verify that wire structs
56 * have the proper alignment/size on all architectures.
58 * DO NOT CHANGE any of the structs, flags, values declared here and used
59 * in released Lustre versions. Some structs may have padding fields that
60 * can be used. Some structs might allow addition at the end (verify this
61 * in the code to ensure that new/old clients that see this larger struct
62 * do not fail, otherwise you need to implement protocol compatibility).
67 #ifndef _LUSTRE_IDL_H_
68 #define _LUSTRE_IDL_H_
70 #include <asm/byteorder.h>
71 #include <linux/types.h>
73 #include <uapi/linux/lnet/lnet-types.h>
74 /* Defn's shared with user-space. */
75 #include <uapi/linux/lustre/lustre_user.h>
76 #include <uapi/linux/lustre/lustre_ver.h>
81 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
82 * FOO_REPLY_PORTAL is for incoming replies on the FOO
83 * FOO_BULK_PORTAL is for incoming bulk on the FOO
86 /* Lustre service names are following the format
87 * service name + MDT + seq name
89 #define LUSTRE_MDT_MAXNAMELEN 80
91 #define CONNMGR_REQUEST_PORTAL 1
92 #define CONNMGR_REPLY_PORTAL 2
93 /*#define OSC_REQUEST_PORTAL 3 */
94 #define OSC_REPLY_PORTAL 4
95 /*#define OSC_BULK_PORTAL 5 */
96 #define OST_IO_PORTAL 6
97 #define OST_CREATE_PORTAL 7
98 #define OST_BULK_PORTAL 8
99 /*#define MDC_REQUEST_PORTAL 9 */
100 #define MDC_REPLY_PORTAL 10
101 /*#define MDC_BULK_PORTAL 11 */
102 #define MDS_REQUEST_PORTAL 12
103 /*#define MDS_REPLY_PORTAL 13 */
104 #define MDS_BULK_PORTAL 14
105 #define LDLM_CB_REQUEST_PORTAL 15
106 #define LDLM_CB_REPLY_PORTAL 16
107 #define LDLM_CANCEL_REQUEST_PORTAL 17
108 #define LDLM_CANCEL_REPLY_PORTAL 18
109 /*#define PTLBD_REQUEST_PORTAL 19 */
110 /*#define PTLBD_REPLY_PORTAL 20 */
111 /*#define PTLBD_BULK_PORTAL 21 */
112 #define MDS_SETATTR_PORTAL 22
113 #define MDS_READPAGE_PORTAL 23
114 #define OUT_PORTAL 24
116 #define MGC_REPLY_PORTAL 25
117 #define MGS_REQUEST_PORTAL 26
118 #define MGS_REPLY_PORTAL 27
119 #define OST_REQUEST_PORTAL 28
120 #define FLD_REQUEST_PORTAL 29
121 #define SEQ_METADATA_PORTAL 30
122 #define SEQ_DATA_PORTAL 31
123 #define SEQ_CONTROLLER_PORTAL 32
124 #define MGS_BULK_PORTAL 33
126 /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com,
131 #define PTL_RPC_MSG_REQUEST 4711
132 #define PTL_RPC_MSG_ERR 4712
133 #define PTL_RPC_MSG_REPLY 4713
135 /* DON'T use swabbed values of MAGIC as magic! */
136 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
137 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
139 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
141 #define PTLRPC_MSG_VERSION 0x00000003
142 #define LUSTRE_VERSION_MASK 0xffff0000
143 #define LUSTRE_OBD_VERSION 0x00010000
144 #define LUSTRE_MDS_VERSION 0x00020000
145 #define LUSTRE_OST_VERSION 0x00030000
146 #define LUSTRE_DLM_VERSION 0x00040000
147 #define LUSTRE_LOG_VERSION 0x00050000
148 #define LUSTRE_MGS_VERSION 0x00060000
151 * Describes a range of sequence, lsr_start is included but lsr_end is
153 * Same structure is used in fld module where lsr_index field holds mdt id
156 struct lu_seq_range {
163 struct lu_seq_range_array {
166 struct lu_seq_range lsra_lsr[0];
169 #define LU_SEQ_RANGE_MDT 0x0
170 #define LU_SEQ_RANGE_OST 0x1
171 #define LU_SEQ_RANGE_ANY 0x3
173 #define LU_SEQ_RANGE_MASK 0x3
175 /** \defgroup lu_fid lu_fid
180 * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat.
181 * Deprecated since HSM and SOM attributes are now stored in separate on-disk
185 LMAC_HSM = 0x00000001,
186 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
187 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
188 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
189 * under /O/<seq>/d<x>.
194 * Masks for all features that should be supported by a Lustre version to
195 * access a specific file.
196 * This information is stored in lustre_mdt_attrs::lma_incompat.
199 LMAI_RELEASED = 0x00000001, /* file is released */
200 LMAI_AGENT = 0x00000002, /* agent inode */
201 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
202 * is on the remote MDT
206 #define LMA_INCOMPAT_SUPP (LMAI_AGENT | LMAI_REMOTE_PARENT)
212 /** LASTID file has zero OID */
213 LUSTRE_FID_LASTID_OID = 0UL,
214 /** initial fid id value */
215 LUSTRE_FID_INIT_OID = 1UL
218 /* copytool uses a 32b bitmask field to encode archive-Ids during register
220 * archive num = 0 => all
221 * archive num from 1 to 32
223 #define LL_HSM_MAX_ARCHIVE (sizeof(__u32) * 8)
226 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
227 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
228 * used for other purposes and not risk collisions with existing inodes.
230 * Different FID Format
231 * http://wiki.old.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs
234 FID_SEQ_OST_MDT0 = 0,
235 FID_SEQ_LLOG = 1, /* unnamed llogs */
237 FID_SEQ_OST_MDT1 = 3,
238 FID_SEQ_OST_MAX = 9, /* Max MDT count before OST_on_FID */
239 FID_SEQ_LLOG_NAME = 10, /* named llogs */
242 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
243 FID_SEQ_IDIF = 0x100000000ULL,
244 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
245 /* Normal FID sequence starts from this value, i.e. 1<<33 */
246 FID_SEQ_START = 0x200000000ULL,
247 /* sequence for local pre-defined FIDs listed in local_oid */
248 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
249 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
250 /* sequence is used for local named objects FIDs generated
251 * by local_object_storage library
253 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
254 /* Because current FLD will only cache the fid sequence, instead
255 * of oid on the client side, if the FID needs to be exposed to
256 * clients sides, it needs to make sure all of fids under one
257 * sequence will be located in one MDT.
259 FID_SEQ_SPECIAL = 0x200000004ULL,
260 FID_SEQ_QUOTA = 0x200000005ULL,
261 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
262 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
263 FID_SEQ_NORMAL = 0x200000400ULL,
264 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
267 #define OBIF_OID_MAX_BITS 32
268 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
269 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
270 #define IDIF_OID_MAX_BITS 48
271 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
272 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
274 /** OID for FID_SEQ_SPECIAL */
276 /* Big Filesystem Lock to serialize rename operations */
277 FID_OID_SPECIAL_BFL = 1UL,
280 /** OID for FID_SEQ_DOT_LUSTRE */
281 enum dot_lustre_oid {
282 FID_OID_DOT_LUSTRE = 1UL,
283 FID_OID_DOT_LUSTRE_OBF = 2UL,
286 /** OID for FID_SEQ_ROOT */
289 FID_OID_ECHO_ROOT = 2UL,
294 /** \defgroup lu_dir lu_dir
299 * Enumeration of possible directory entry attributes.
301 * Attributes follow directory entry header in the order they appear in this
304 enum lu_dirent_attrs {
307 LUDA_64BITHASH = 0x0004,
311 * Layout of readdir pages, as transmitted on wire.
314 /** valid if LUDA_FID is set. */
315 struct lu_fid lde_fid;
316 /** a unique entry identifier: a hash or an offset. */
318 /** total record length, including all attributes. */
322 /** optional variable size attributes following this entry.
323 * taken from enum lu_dirent_attrs.
326 /** name is followed by the attributes indicated in ->ldp_attrs, in
327 * their natural order. After the last attribute, padding bytes are
328 * added to make ->lde_reclen a multiple of 8.
334 * Definitions of optional directory entry attributes formats.
336 * Individual attributes do not have their length encoded in a generic way. It
337 * is assumed that consumer of an attribute knows its format. This means that
338 * it is impossible to skip over an unknown attribute, except by skipping over all
339 * remaining attributes (by using ->lde_reclen), which is not too
340 * constraining, because new server versions will append new attributes at
341 * the end of an entry.
345 * Fid directory attribute: a fid of an object referenced by the entry. This
346 * will be almost always requested by the client and supplied by the server.
348 * Aligned to 8 bytes.
350 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
355 * Aligned to 2 bytes.
366 #define IFTODT(type) (((type) & S_IFMT) >> IFSHIFT)
369 #define DTTOIF(dirtype) ((dirtype) << IFSHIFT)
373 __le64 ldp_hash_start;
377 struct lu_dirent ldp_entries[0];
380 enum lu_dirpage_flags {
382 * dirpage contains no entry.
386 * last entry's lde_hash equals ldp_hash_end.
391 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
393 if (__le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
396 return dp->ldp_entries;
399 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
401 struct lu_dirent *next;
403 if (__le16_to_cpu(ent->lde_reclen) != 0)
404 next = ((void *)ent) + __le16_to_cpu(ent->lde_reclen);
411 static inline size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
415 if (attr & LUDA_TYPE) {
416 const size_t align = sizeof(struct luda_type) - 1;
418 size = (sizeof(struct lu_dirent) + namelen + align) & ~align;
419 size += sizeof(struct luda_type);
421 size = sizeof(struct lu_dirent) + namelen;
424 return (size + 7) & ~7;
427 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
430 * MDS_READPAGE page size
432 * This is the directory page size packed in MDS_READPAGE RPC.
433 * It's different than PAGE_SIZE because the client needs to
434 * access the struct lu_dirpage header packed at the beginning of
435 * the "page" and without this there isn't any way to know find the
436 * lu_dirpage header is if client and server PAGE_SIZE differ.
438 #define LU_PAGE_SHIFT 12
439 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
440 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
442 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
446 struct lustre_handle {
450 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
452 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
454 return lh->cookie != 0ull;
457 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
458 const struct lustre_handle *lh2)
460 return lh1->cookie == lh2->cookie;
463 static inline void lustre_handle_copy(struct lustre_handle *tgt,
464 const struct lustre_handle *src)
466 tgt->cookie = src->cookie;
469 /* flags for lm_flags */
470 #define MSGHDR_AT_SUPPORT 0x1
471 #define MSGHDR_CKSUM_INCOMPAT18 0x2
473 #define lustre_msg lustre_msg_v2
474 /* we depend on this structure to be 8-byte aligned */
475 /* this type is only endian-adjusted in lustre_unpack_msg() */
476 struct lustre_msg_v2 {
488 /* without gss, ptlrpc_body is put at the first buffer. */
489 #define PTLRPC_NUM_VERSIONS 4
491 struct ptlrpc_body_v3 {
492 struct lustre_handle pb_handle;
497 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
498 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
501 __u64 pb_last_committed;
506 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
507 __u32 pb_service_time; /* for rep, actual service time */
510 /* VBR: pre-versions */
511 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
512 __u64 pb_mbits; /**< match bits for bulk request */
513 /* padding for future needs */
514 __u64 pb_padding64_0;
515 __u64 pb_padding64_1;
516 __u64 pb_padding64_2;
517 char pb_jobid[LUSTRE_JOBID_SIZE];
520 #define ptlrpc_body ptlrpc_body_v3
522 struct ptlrpc_body_v2 {
523 struct lustre_handle pb_handle;
528 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
529 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
532 __u64 pb_last_committed;
537 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
538 __u32 pb_service_time; /* for rep, actual service time, also used for
543 /* VBR: pre-versions */
544 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
545 __u64 pb_mbits; /**< unused in V2 */
546 /* padding for future needs */
547 __u64 pb_padding64_0;
548 __u64 pb_padding64_1;
549 __u64 pb_padding64_2;
552 /* message body offset for lustre_msg_v2 */
553 /* ptlrpc body offset in all request/reply messages */
554 #define MSG_PTLRPC_BODY_OFF 0
556 /* normal request/reply message record offset */
557 #define REQ_REC_OFF 1
558 #define REPLY_REC_OFF 1
560 /* ldlm request message body offset */
561 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
562 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
564 /* ldlm intent lock message body offset */
565 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
566 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
568 /* ldlm reply message body offset */
569 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
570 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
572 /** only use in req->rq_{req,rep}_swab_mask */
573 #define MSG_PTLRPC_HEADER_OFF 31
575 /* Flags that are operation-specific go in the top 16 bits. */
576 #define MSG_OP_FLAG_MASK 0xffff0000
577 #define MSG_OP_FLAG_SHIFT 16
579 /* Flags that apply to all requests are in the bottom 16 bits */
580 #define MSG_GEN_FLAG_MASK 0x0000ffff
581 #define MSG_LAST_REPLAY 0x0001
582 #define MSG_RESENT 0x0002
583 #define MSG_REPLAY 0x0004
584 /* #define MSG_AT_SUPPORT 0x0008
585 * This was used in early prototypes of adaptive timeouts, and while there
586 * shouldn't be any users of that code there also isn't a need for using this
587 * bits. Defer usage until at least 1.10 to avoid potential conflict.
589 #define MSG_DELAY_REPLAY 0x0010
590 #define MSG_VERSION_REPLAY 0x0020
591 #define MSG_REQ_REPLAY_DONE 0x0040
592 #define MSG_LOCK_REPLAY_DONE 0x0080
595 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
598 #define MSG_CONNECT_RECOVERING 0x00000001
599 #define MSG_CONNECT_RECONNECT 0x00000002
600 #define MSG_CONNECT_REPLAYABLE 0x00000004
601 /*#define MSG_CONNECT_PEER 0x8 */
602 #define MSG_CONNECT_LIBCLIENT 0x00000010
603 #define MSG_CONNECT_INITIAL 0x00000020
604 #define MSG_CONNECT_ASYNC 0x00000040
605 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
606 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
609 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
610 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
611 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
612 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
613 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
614 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
615 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
616 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
617 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
618 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
619 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
620 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
621 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
622 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
623 *We do not support JOIN FILE
624 *anymore, reserve this flags
625 *just for preventing such bit
628 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
629 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
630 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
631 * in production. Removed in
632 * 2.9. Keep this flag to
635 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
636 * never used in production.
637 * Removed in 2.9. Keep this
638 * flag to avoid reuse
640 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
641 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
642 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
643 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
644 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
645 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
646 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
647 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
648 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
649 #define OBD_CONNECT_REAL 0x8000000ULL /* obsolete since 2.8 */
650 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
651 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
652 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
653 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
654 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
655 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
656 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
657 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
658 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
659 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
660 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
663 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
664 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
665 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
666 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
667 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
670 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
671 * finer space reservation
673 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
674 * policy and 2.x server
676 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
677 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
678 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
679 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
680 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
681 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* flock deadlock detection */
682 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/*create stripe disposition*/
683 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
686 #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */
687 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */
688 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify
691 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL/* striped DNE dir */
692 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
693 #define OBD_CONNECT_LOCK_AHEAD 0x1000000000000000ULL /* lock ahead */
694 /** bulk matchbits is sent within ptlrpc_body */
695 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL
696 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
697 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
700 * Please DO NOT add flag values here before first ensuring that this same
701 * flag value is not in use on some other branch. Please clear any such
702 * changes with senior engineers before starting to use a new flag. Then,
703 * submit a small patch against EVERY branch that ONLY adds the new flag,
704 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
705 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
706 * can be approved and landed easily to reserve the flag for future use.
709 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
710 * connection. It is a temporary bug fix for Imperative Recovery interop
711 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
712 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644.
714 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
716 #define OCD_HAS_FLAG(ocd, flg) \
717 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
719 /* Features required for this version of the client to work with server */
720 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
723 /* This structure is used for both request and reply.
725 * If we eventually have separate connect data for different types, which we
726 * almost certainly will, then perhaps we stick a union in here.
728 struct obd_connect_data {
729 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
730 __u32 ocd_version; /* lustre release version number */
731 __u32 ocd_grant; /* initial cache grant amount (bytes) */
732 __u32 ocd_index; /* LOV index to connect to */
733 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
734 __u64 ocd_ibits_known; /* inode bits this client understands */
735 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
736 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
737 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
738 __u32 ocd_unused; /* also fix lustre_swab_connect */
739 __u64 ocd_transno; /* first transno from client to be replayed */
740 __u32 ocd_group; /* MDS group on OST */
741 __u32 ocd_cksum_types; /* supported checksum algorithms */
742 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
743 __u32 ocd_instance; /* instance # of this target */
744 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
745 /* Fields after ocd_maxbytes are only accessible by the receiver
746 * if the corresponding flag in ocd_connect_flags is set. Accessing
747 * any field after ocd_maxbytes on the receiver without a valid flag
748 * may result in out-of-bound memory access and kernel oops.
750 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
751 __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */
752 __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */
753 __u64 ocd_connect_flags2;
754 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
755 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
756 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
757 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
758 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
759 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
760 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
761 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
762 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
763 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
764 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
765 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
766 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
770 * Please DO NOT use any fields here before first ensuring that this same
771 * field is not in use on some other branch. Please clear any such changes
772 * with senior engineers before starting to use a new field. Then, submit
773 * a small patch against EVERY branch that ONLY adds the new field along with
774 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
775 * reserve the flag for future use.
779 * Supported checksum algorithms. Up to 32 checksum types are supported.
780 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
781 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
782 * algorithm and also the OBD_FL_CKSUM* flags.
785 OBD_CKSUM_CRC32 = 0x00000001,
786 OBD_CKSUM_ADLER = 0x00000002,
787 OBD_CKSUM_CRC32C = 0x00000004,
791 * OST requests: OBDO & OBD request records
796 OST_REPLY = 0, /* reply ? */
812 OST_QUOTACHECK = 18, /* not used since 2.4 */
814 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
817 #define OST_FIRST_OPC OST_REPLY
820 OBD_FL_INLINEDATA = 0x00000001,
821 OBD_FL_OBDMDEXISTS = 0x00000002,
822 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
823 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
824 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
825 OBD_FL_RECREATE_OBJS = 0x00000020, /* recreate missing obj */
826 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
827 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
828 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
829 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
830 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
831 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
832 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
833 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
834 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
835 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
836 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
837 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
838 * XXX: obsoleted - reserved for old
839 * clients prior than 2.2
841 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
842 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
843 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
844 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
846 /* Note that while these checksum values are currently separate bits,
847 * in 2.x we can actually allow all values from 1-31 if we wanted.
849 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
852 /* mask for local-only flag, which won't be sent over network */
853 OBD_FL_LOCAL_MASK = 0xF0000000,
857 * All LOV EA magics should have the same postfix, if some new version
858 * Lustre instroduces new LOV EA magic, then when down-grade to an old
859 * Lustre, even though the old version system does not recognizes such
860 * new magic, it still can distinguish the corrupted cases by checking
861 * the magic's postfix.
863 #define LOV_MAGIC_MAGIC 0x0BD0
864 #define LOV_MAGIC_MASK 0xFFFF
866 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
867 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
868 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
869 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
870 /* reserved for specifying OSTs */
871 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
872 #define LOV_MAGIC LOV_MAGIC_V1
875 * magic for fully defined striping
876 * the idea is that we should have different magics for striping "hints"
877 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
878 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
879 * we can't just change it w/o long way preparation, but we still need a
880 * mechanism to allow LOD to differentiate hint versus ready striping.
881 * so, at the moment we do a trick: MDT knows what to expect from request
882 * depending on the case (replay uses ready striping, non-replay req uses
883 * hints), so MDT replaces magic with appropriate one and now LOD can
884 * easily understand what's inside -bzzz
886 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
887 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
889 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
890 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
892 #define lov_ost_data lov_ost_data_v1
893 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
894 struct ost_id l_ost_oi; /* OST object ID */
895 __u32 l_ost_gen; /* generation of this l_ost_idx */
896 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
899 #define lov_mds_md lov_mds_md_v1
900 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
901 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
902 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
903 struct ost_id lmm_oi; /* LOV object ID */
904 __u32 lmm_stripe_size; /* size of stripe in bytes */
905 /* lmm_stripe_count used to be __u32 */
906 __u16 lmm_stripe_count; /* num stripes in use for this object */
907 __u16 lmm_layout_gen; /* layout generation number */
908 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
911 #define MAX_MD_SIZE \
912 (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
913 #define MIN_MD_SIZE \
914 (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
916 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
917 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
918 #define XATTR_USER_PREFIX "user."
919 #define XATTR_TRUSTED_PREFIX "trusted."
920 #define XATTR_SECURITY_PREFIX "security."
921 #define XATTR_LUSTRE_PREFIX "lustre."
923 #define XATTR_NAME_LOV "trusted.lov"
924 #define XATTR_NAME_LMA "trusted.lma"
925 #define XATTR_NAME_LMV "trusted.lmv"
926 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
927 #define XATTR_NAME_LINK "trusted.link"
928 #define XATTR_NAME_FID "trusted.fid"
929 #define XATTR_NAME_VERSION "trusted.version"
930 #define XATTR_NAME_SOM "trusted.som"
931 #define XATTR_NAME_HSM "trusted.hsm"
932 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_namespace"
934 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
935 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
936 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
937 struct ost_id lmm_oi; /* LOV object ID */
938 __u32 lmm_stripe_size; /* size of stripe in bytes */
939 /* lmm_stripe_count used to be __u32 */
940 __u16 lmm_stripe_count; /* num stripes in use for this object */
941 __u16 lmm_layout_gen; /* layout generation number */
942 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
943 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
946 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
948 if (lmm_magic == LOV_MAGIC_V3)
949 return sizeof(struct lov_mds_md_v3) +
950 stripes * sizeof(struct lov_ost_data_v1);
952 return sizeof(struct lov_mds_md_v1) +
953 stripes * sizeof(struct lov_ost_data_v1);
957 lov_mds_md_max_stripe_count(size_t buf_size, __u32 lmm_magic)
961 struct lov_mds_md_v1 lmm;
963 if (buf_size < sizeof(lmm))
966 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
969 struct lov_mds_md_v3 lmm;
971 if (buf_size < sizeof(lmm))
974 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
981 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
982 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
983 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
984 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
985 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
986 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
987 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
988 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
989 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
990 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
991 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
992 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
993 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
994 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
995 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
996 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
997 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
998 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
999 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1000 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1001 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1002 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1003 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1004 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1005 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1006 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1007 /* ->mds if epoch opens or closes
1009 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1010 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1011 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1012 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1013 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1015 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1016 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1017 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1018 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1020 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1021 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1022 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1023 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1024 /* OBD_MD_FLRMTPERM (0x0000010000000000ULL) remote perm, obsolete */
1025 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1026 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1027 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1028 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1029 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1030 * under lock; for xattr
1031 * requests means the
1032 * client holds the lock
1034 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1036 /* OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) lfs lsetfacl, obsolete */
1037 /* OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) lfs lgetfacl, obsolete */
1038 /* OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) lfs rsetfacl, obsolete */
1039 /* OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) lfs rgetfacl, obsolete */
1041 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1042 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1046 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1048 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1049 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1050 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1051 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1052 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1054 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1056 /* don't forget obdo_fid which is way down at the bottom so it can
1057 * come after the definition of llog_cookie
1062 HSS_CLEARMASK = 0x02,
1063 HSS_ARCHIVE_ID = 0x04,
1066 struct hsm_state_set {
1068 __u32 hss_archive_id;
1070 __u64 hss_clearmask;
1073 /* ost_body.data values for OST_BRW */
1075 #define OBD_BRW_READ 0x01
1076 #define OBD_BRW_WRITE 0x02
1077 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1078 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1079 * transfer and is not accounted in
1082 #define OBD_BRW_CHECK 0x10
1083 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1084 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1085 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1086 #define OBD_BRW_NOQUOTA 0x100
1087 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1088 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1089 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1090 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1091 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1092 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1093 * that the client is running low on
1094 * space for unstable pages; asking
1095 * it to sync quickly
1098 #define OBD_OBJECT_EOF LUSTRE_EOF
1100 #define OST_MIN_PRECREATE 32
1101 #define OST_MAX_PRECREATE 20000
1104 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1105 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1106 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1107 * high 16 bits in 2.4 and later
1109 __u32 ioo_bufcnt; /* number of niobufs for this object */
1113 * NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1114 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1115 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits.
1117 #define IOOBJ_MAX_BRW_BITS 16
1118 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1119 #define ioobj_max_brw_set(ioo, num) \
1120 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1122 /* multiple of 8 bytes => can array */
1123 struct niobuf_remote {
1129 /* lock value block communicated between the filter and llite */
1131 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1132 * negative, i.e. because ((MASK + rc) & MASK) != MASK.
1134 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1135 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1136 #define OST_LVB_IS_ERR(blocks) \
1137 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1138 #define OST_LVB_SET_ERR(blocks, rc) \
1139 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1140 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1163 * lquota data structures
1166 /* The lquota_id structure is a union of all the possible identifier types that
1167 * can be used with quota, this includes:
1170 * - a FID which can be used for per-directory quota in the future
1173 struct lu_fid qid_fid; /* FID for per-directory quota */
1174 __u64 qid_uid; /* user identifier */
1175 __u64 qid_gid; /* group identifier */
1178 /* quotactl management */
1179 struct obd_quotactl {
1181 __u32 qc_type; /* see Q_* flag below */
1184 struct obd_dqinfo qc_dqinfo;
1185 struct obd_dqblk qc_dqblk;
1188 #define Q_COPY(out, in, member) (out)->member = (in)->member
1190 #define QCTL_COPY(out, in) \
1192 Q_COPY(out, in, qc_cmd); \
1193 Q_COPY(out, in, qc_type); \
1194 Q_COPY(out, in, qc_id); \
1195 Q_COPY(out, in, qc_stat); \
1196 Q_COPY(out, in, qc_dqinfo); \
1197 Q_COPY(out, in, qc_dqblk); \
1200 /* Data structures associated with the quota locks */
1202 /* Glimpse descriptor used for the index & per-ID quota locks */
1203 struct ldlm_gl_lquota_desc {
1204 union lquota_id gl_id; /* quota ID subject to the glimpse */
1205 __u64 gl_flags; /* see LQUOTA_FL* below */
1206 __u64 gl_ver; /* new index version */
1207 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1208 __u64 gl_softlimit; /* new softlimit */
1213 /* quota glimpse flags */
1214 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1216 /* LVB used with quota (global and per-ID) locks */
1218 __u64 lvb_flags; /* see LQUOTA_FL* above */
1219 __u64 lvb_id_may_rel; /* space that might be released later */
1220 __u64 lvb_id_rel; /* space released by the slave for this ID */
1221 __u64 lvb_id_qunit; /* current qunit value */
1231 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1240 MDS_GETATTR_NAME = 34,
1245 MDS_DISCONNECT = 39,
1248 MDS_PIN = 42, /* obsolete, never used in a release */
1249 MDS_UNPIN = 43, /* obsolete, never used in a release */
1251 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1253 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1256 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1258 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1260 MDS_HSM_STATE_GET = 54,
1261 MDS_HSM_STATE_SET = 55,
1262 MDS_HSM_ACTION = 56,
1263 MDS_HSM_PROGRESS = 57,
1264 MDS_HSM_REQUEST = 58,
1265 MDS_HSM_CT_REGISTER = 59,
1266 MDS_HSM_CT_UNREGISTER = 60,
1267 MDS_SWAP_LAYOUTS = 61,
1271 #define MDS_FIRST_OPC MDS_GETATTR
1277 enum mdt_reint_cmd {
1290 /* the disposition of the intent outlines what was executed */
1291 #define DISP_IT_EXECD 0x00000001
1292 #define DISP_LOOKUP_EXECD 0x00000002
1293 #define DISP_LOOKUP_NEG 0x00000004
1294 #define DISP_LOOKUP_POS 0x00000008
1295 #define DISP_OPEN_CREATE 0x00000010
1296 #define DISP_OPEN_OPEN 0x00000020
1297 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1298 #define DISP_ENQ_OPEN_REF 0x00800000
1299 #define DISP_ENQ_CREATE_REF 0x01000000
1300 #define DISP_OPEN_LOCK 0x02000000
1301 #define DISP_OPEN_LEASE 0x04000000
1302 #define DISP_OPEN_STRIPE 0x08000000
1303 #define DISP_OPEN_DENY 0x10000000
1305 /* INODE LOCK PARTS */
1306 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
1307 * was used to protect permission (mode,
1308 * owner, group etc) before 2.4.
1310 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1311 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1312 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1314 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
1315 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
1316 * Because for remote directories(in DNE), these locks will be granted by
1317 * different MDTs(different ldlm namespace).
1319 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
1320 * For Remote directory, the master MDT, where the remote directory is, will
1321 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
1322 * will grant LOOKUP_LOCK.
1324 #define MDS_INODELOCK_PERM 0x000010
1325 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
1327 #define MDS_INODELOCK_MAXSHIFT 5
1328 /* This FULL lock is useful to take on unlink sort of operations */
1329 #define MDS_INODELOCK_FULL ((1 << (MDS_INODELOCK_MAXSHIFT + 1)) - 1)
1331 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1332 * but was moved into name[1] along with the OID to avoid consuming the
1333 * name[2,3] fields that need to be used for the quota id (also a FID).
1336 LUSTRE_RES_ID_SEQ_OFF = 0,
1337 LUSTRE_RES_ID_VER_OID_OFF = 1,
1338 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1339 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1340 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1341 LUSTRE_RES_ID_HSH_OFF = 3
1344 #define MDS_STATUS_CONN 1
1345 #define MDS_STATUS_LOV 2
1347 /* these should be identical to their EXT4_*_FL counterparts, they are
1348 * redefined here only to avoid dragging in fs/ext4/ext4.h
1350 #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */
1351 #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */
1352 #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */
1353 #define LUSTRE_NODUMP_FL 0x00000040 /* do not dump file */
1354 #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */
1355 #define LUSTRE_INDEX_FL 0x00001000 /* hash-indexed directory */
1356 #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1357 #define LUSTRE_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
1358 #define LUSTRE_DIRECTIO_FL 0x00100000 /* Use direct i/o */
1359 #define LUSTRE_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */
1361 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1362 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1363 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1364 * the S_* flags are kernel-internal values that change between kernel
1365 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1366 * See b=16526 for a full history.
1368 static inline int ll_ext_to_inode_flags(int flags)
1370 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1371 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1372 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1373 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1374 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0));
1377 static inline int ll_inode_to_ext_flags(int iflags)
1379 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1380 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1381 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1382 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1383 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0));
1386 /* 64 possible states */
1387 enum md_transient_state {
1388 MS_RESTORE = (1 << 0), /* restore is running */
1392 struct lu_fid mbo_fid1;
1393 struct lu_fid mbo_fid2;
1394 struct lustre_handle mbo_handle;
1396 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1400 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1402 __u64 mbo_t_state; /* transient file state defined in
1403 * enum md_transient_state
1404 * was "ino" until 2.4.0
1408 __u32 mbo_capability;
1412 __u32 mbo_flags; /* LUSTRE_*_FL file attributes */
1414 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1415 __u32 mbo_unused2; /* was "generation" until 2.4.0 */
1417 __u32 mbo_eadatasize;
1419 __u32 mbo_max_mdsize;
1420 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1421 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1422 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1423 __u32 mbo_padding_5; /* also fix lustre_swab_mdt_body */
1424 __u64 mbo_padding_6;
1425 __u64 mbo_padding_7;
1426 __u64 mbo_padding_8;
1427 __u64 mbo_padding_9;
1428 __u64 mbo_padding_10;
1431 struct mdt_ioepoch {
1432 struct lustre_handle mio_handle;
1433 __u64 mio_unused1; /* was ioepoch */
1434 __u32 mio_unused2; /* was flags */
1438 /* permissions for md_perm.mp_perm */
1440 CFS_SETUID_PERM = 0x01,
1441 CFS_SETGID_PERM = 0x02,
1442 CFS_SETGRP_PERM = 0x04,
1445 struct mdt_rec_setattr {
1455 __u32 sa_padding_1_h;
1456 struct lu_fid sa_fid;
1465 __u32 sa_attr_flags;
1467 __u32 sa_bias; /* some operation flags */
1474 * Attribute flags used in mdt_rec_setattr::sa_valid.
1475 * The kernel's #defines for ATTR_* should not be used over the network
1476 * since the client and MDS may run different kernels (see bug 13828)
1477 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1479 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1480 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1481 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1482 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1483 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1484 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1485 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1486 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1487 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1488 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1489 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1490 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1491 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1492 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1493 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path,
1496 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1498 #define MDS_FMODE_CLOSED 00000000
1499 #define MDS_FMODE_EXEC 00000004
1500 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1501 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1502 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1504 #define MDS_OPEN_CREATED 00000010
1505 #define MDS_OPEN_CROSS 00000020
1507 #define MDS_OPEN_CREAT 00000100
1508 #define MDS_OPEN_EXCL 00000200
1509 #define MDS_OPEN_TRUNC 00001000
1510 #define MDS_OPEN_APPEND 00002000
1511 #define MDS_OPEN_SYNC 00010000
1512 #define MDS_OPEN_DIRECTORY 00200000
1514 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1515 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1516 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1517 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1518 * We do not support JOIN FILE
1519 * anymore, reserve this flags
1520 * just for preventing such bit
1524 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1525 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1526 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1527 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1528 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1530 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1532 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1533 * delegation, succeed if it's not
1534 * being opened with conflict mode.
1536 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1538 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1539 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1540 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1544 MDS_CHECK_SPLIT = 1 << 0,
1545 MDS_CROSS_REF = 1 << 1,
1546 MDS_VTX_BYPASS = 1 << 2,
1547 MDS_PERM_BYPASS = 1 << 3,
1548 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1549 MDS_QUOTA_IGNORE = 1 << 5,
1550 MDS_CLOSE_CLEANUP = 1 << 6,
1551 MDS_KEEP_ORPHAN = 1 << 7,
1552 MDS_RECOV_OPEN = 1 << 8,
1553 MDS_DATA_MODIFIED = 1 << 9,
1554 MDS_CREATE_VOLATILE = 1 << 10,
1555 MDS_OWNEROVERRIDE = 1 << 11,
1556 MDS_HSM_RELEASE = 1 << 12,
1557 MDS_RENAME_MIGRATE = 1 << 13,
1558 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1561 /* instance of mdt_reint_rec */
1562 struct mdt_rec_create {
1570 __u32 cr_suppgid1_h;
1572 __u32 cr_suppgid2_h;
1573 struct lu_fid cr_fid1;
1574 struct lu_fid cr_fid2;
1575 struct lustre_handle cr_old_handle; /* handle in case of open replay */
1579 __u64 cr_padding_1; /* rr_blocks */
1582 /* use of helpers set/get_mrc_cr_flags() is needed to access
1583 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
1584 * extend cr_flags size without breaking 1.8 compat
1586 __u32 cr_flags_l; /* for use with open, low 32 bits */
1587 __u32 cr_flags_h; /* for use with open, high 32 bits */
1588 __u32 cr_umask; /* umask for create */
1589 __u32 cr_padding_4; /* rr_padding_4 */
1592 /* instance of mdt_reint_rec */
1593 struct mdt_rec_link {
1601 __u32 lk_suppgid1_h;
1603 __u32 lk_suppgid2_h;
1604 struct lu_fid lk_fid1;
1605 struct lu_fid lk_fid2;
1607 __u64 lk_padding_1; /* rr_atime */
1608 __u64 lk_padding_2; /* rr_ctime */
1609 __u64 lk_padding_3; /* rr_size */
1610 __u64 lk_padding_4; /* rr_blocks */
1612 __u32 lk_padding_5; /* rr_mode */
1613 __u32 lk_padding_6; /* rr_flags */
1614 __u32 lk_padding_7; /* rr_padding_2 */
1615 __u32 lk_padding_8; /* rr_padding_3 */
1616 __u32 lk_padding_9; /* rr_padding_4 */
1619 /* instance of mdt_reint_rec */
1620 struct mdt_rec_unlink {
1628 __u32 ul_suppgid1_h;
1630 __u32 ul_suppgid2_h;
1631 struct lu_fid ul_fid1;
1632 struct lu_fid ul_fid2;
1634 __u64 ul_padding_2; /* rr_atime */
1635 __u64 ul_padding_3; /* rr_ctime */
1636 __u64 ul_padding_4; /* rr_size */
1637 __u64 ul_padding_5; /* rr_blocks */
1640 __u32 ul_padding_6; /* rr_flags */
1641 __u32 ul_padding_7; /* rr_padding_2 */
1642 __u32 ul_padding_8; /* rr_padding_3 */
1643 __u32 ul_padding_9; /* rr_padding_4 */
1646 /* instance of mdt_reint_rec */
1647 struct mdt_rec_rename {
1655 __u32 rn_suppgid1_h;
1657 __u32 rn_suppgid2_h;
1658 struct lu_fid rn_fid1;
1659 struct lu_fid rn_fid2;
1661 __u64 rn_padding_1; /* rr_atime */
1662 __u64 rn_padding_2; /* rr_ctime */
1663 __u64 rn_padding_3; /* rr_size */
1664 __u64 rn_padding_4; /* rr_blocks */
1665 __u32 rn_bias; /* some operation flags */
1666 __u32 rn_mode; /* cross-ref rename has mode */
1667 __u32 rn_padding_5; /* rr_flags */
1668 __u32 rn_padding_6; /* rr_padding_2 */
1669 __u32 rn_padding_7; /* rr_padding_3 */
1670 __u32 rn_padding_8; /* rr_padding_4 */
1673 /* instance of mdt_reint_rec */
1674 struct mdt_rec_setxattr {
1682 __u32 sx_suppgid1_h;
1684 __u32 sx_suppgid2_h;
1685 struct lu_fid sx_fid;
1686 __u64 sx_padding_1; /* These three are rr_fid2 */
1691 __u64 sx_padding_5; /* rr_ctime */
1692 __u64 sx_padding_6; /* rr_size */
1693 __u64 sx_padding_7; /* rr_blocks */
1696 __u32 sx_padding_8; /* rr_flags */
1697 __u32 sx_padding_9; /* rr_padding_2 */
1698 __u32 sx_padding_10; /* rr_padding_3 */
1699 __u32 sx_padding_11; /* rr_padding_4 */
1703 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
1704 * Do NOT change the size of various members, otherwise the value
1705 * will be broken in lustre_swab_mdt_rec_reint().
1707 * If you add new members in other mdt_reint_xxx structures and need to use the
1708 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
1710 struct mdt_rec_reint {
1718 __u32 rr_suppgid1_h;
1720 __u32 rr_suppgid2_h;
1721 struct lu_fid rr_fid1;
1722 struct lu_fid rr_fid2;
1733 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1736 /* lmv structures */
1738 __u32 ld_tgt_count; /* how many MDS's */
1739 __u32 ld_active_tgt_count; /* how many active */
1740 __u32 ld_default_stripe_count; /* how many objects are used */
1741 __u32 ld_pattern; /* default hash pattern */
1742 __u64 ld_default_hash_size;
1743 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
1744 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
1745 __u32 ld_qos_maxage; /* in second */
1746 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
1747 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
1748 struct obd_uuid ld_uuid;
1751 /* LMV layout EA, and it will be stored both in master and slave object */
1752 struct lmv_mds_md_v1 {
1754 __u32 lmv_stripe_count;
1755 __u32 lmv_master_mdt_index; /* On master object, it is master
1756 * MDT index, on slave object, it
1757 * is stripe index of the slave obj
1759 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
1760 * which hash function to be used,
1761 * Note: only lower 16 bits is being
1762 * used for now. Higher 16 bits will
1763 * be used to mark the object status,
1764 * for example migrating or dead.
1766 __u32 lmv_layout_version; /* Used for directory restriping */
1770 char lmv_pool_name[LOV_MAXPOOLNAME + 1];/* pool name */
1771 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
1774 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
1775 #define LMV_MAGIC LMV_MAGIC_V1
1777 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
1778 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
1781 *Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1782 * and the higher part will be the flag to indicate the status of object,
1783 * for example the object is being migrated. And the hash function
1784 * might be interpreted differently with different flags.
1786 #define LMV_HASH_TYPE_MASK 0x0000ffff
1788 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1789 #define LMV_HASH_FLAG_DEAD 0x40000000
1792 * The FNV-1a hash algorithm is as follows:
1793 * hash = FNV_offset_basis
1794 * for each octet_of_data to be hashed
1795 * hash = hash XOR octet_of_data
1796 * hash = hash × FNV_prime
1798 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
1800 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
1801 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
1803 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
1804 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
1805 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, size_t size)
1807 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
1808 const unsigned char *p = buf;
1811 for (i = 0; i < size; i++) {
1813 hash *= LUSTRE_FNV_1A_64_PRIME;
1821 struct lmv_mds_md_v1 lmv_md_v1;
1822 struct lmv_user_md lmv_user_md;
1825 static inline ssize_t lmv_mds_md_size(int stripe_count, unsigned int lmm_magic)
1827 ssize_t len = -EINVAL;
1829 switch (lmm_magic) {
1830 case LMV_MAGIC_V1: {
1831 struct lmv_mds_md_v1 *lmm1;
1833 len = sizeof(*lmm1);
1834 len += stripe_count * sizeof(lmm1->lmv_stripe_fids[0]);
1842 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
1844 switch (__le32_to_cpu(lmm->lmv_magic)) {
1846 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
1847 case LMV_USER_MAGIC:
1848 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
1858 FLD_FIRST_OPC = FLD_QUERY
1864 SEQ_FIRST_OPC = SEQ_QUERY
1868 SEQ_ALLOC_SUPER = 0,
1879 * LOV data structures
1882 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1883 /* The size of the buffer the lov/mdc reserves for the
1884 * array of UUIDs returned by the MDS. With the current
1885 * protocol, this will limit the max number of OSTs per LOV
1888 #define LOV_DESC_MAGIC 0xB0CCDE5C
1889 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
1890 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
1892 /* LOV settings descriptor (should only contain static info) */
1894 __u32 ld_tgt_count; /* how many OBD's */
1895 __u32 ld_active_tgt_count; /* how many active */
1896 __u32 ld_default_stripe_count; /* how many objects are used */
1897 __u32 ld_pattern; /* default PATTERN_RAID0 */
1898 __u64 ld_default_stripe_size; /* in bytes */
1899 __u64 ld_default_stripe_offset; /* in bytes */
1900 __u32 ld_padding_0; /* unused */
1901 __u32 ld_qos_maxage; /* in second */
1902 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1903 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1904 struct obd_uuid ld_uuid;
1907 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1912 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1917 LDLM_BL_CALLBACK = 104,
1918 LDLM_CP_CALLBACK = 105,
1919 LDLM_GL_CALLBACK = 106,
1920 LDLM_SET_INFO = 107,
1923 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1925 #define RES_NAME_SIZE 4
1926 struct ldlm_res_id {
1927 __u64 name[RES_NAME_SIZE];
1930 #define DLDLMRES "[%#llx:%#llx:%#llx].%llx"
1931 #define PLDLMRES(res) (res)->lr_name.name[0], (res)->lr_name.name[1], \
1932 (res)->lr_name.name[2], (res)->lr_name.name[3]
1948 #define LCK_MODE_NUM 8
1958 #define LDLM_MIN_TYPE LDLM_PLAIN
1960 struct ldlm_extent {
1966 struct ldlm_inodebits {
1970 struct ldlm_flock_wire {
1978 /* it's important that the fields of the ldlm_extent structure match
1979 * the first fields of the ldlm_flock structure because there is only
1980 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1981 * this ever changes we will need to swab the union differently based
1982 * on the resource type.
1985 union ldlm_wire_policy_data {
1986 struct ldlm_extent l_extent;
1987 struct ldlm_flock_wire l_flock;
1988 struct ldlm_inodebits l_inodebits;
1991 union ldlm_gl_desc {
1992 struct ldlm_gl_lquota_desc lquota_desc;
1995 enum ldlm_intent_flags {
1996 IT_OPEN = 0x00000001,
1997 IT_CREAT = 0x00000002,
1998 IT_OPEN_CREAT = 0x00000003,
1999 IT_READDIR = 0x00000004,
2000 IT_GETATTR = 0x00000008,
2001 IT_LOOKUP = 0x00000010,
2002 IT_UNLINK = 0x00000020,
2003 IT_TRUNC = 0x00000040,
2004 IT_GETXATTR = 0x00000080,
2005 IT_EXEC = 0x00000100,
2006 IT_PIN = 0x00000200,
2007 IT_LAYOUT = 0x00000400,
2008 IT_QUOTA_DQACQ = 0x00000800,
2009 IT_QUOTA_CONN = 0x00001000,
2010 IT_SETXATTR = 0x00002000,
2013 struct ldlm_intent {
2017 struct ldlm_resource_desc {
2018 enum ldlm_type lr_type;
2019 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
2020 struct ldlm_res_id lr_name;
2023 struct ldlm_lock_desc {
2024 struct ldlm_resource_desc l_resource;
2025 enum ldlm_mode l_req_mode;
2026 enum ldlm_mode l_granted_mode;
2027 union ldlm_wire_policy_data l_policy_data;
2030 #define LDLM_LOCKREQ_HANDLES 2
2031 #define LDLM_ENQUEUE_CANCEL_OFF 1
2033 struct ldlm_request {
2036 struct ldlm_lock_desc lock_desc;
2037 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2042 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2043 struct ldlm_lock_desc lock_desc;
2044 struct lustre_handle lock_handle;
2045 __u64 lock_policy_res1;
2046 __u64 lock_policy_res2;
2049 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2050 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2053 * Opcodes for mountconf (mgs and mgc)
2058 MGS_EXCEPTION, /* node died, etc. */
2059 MGS_TARGET_REG, /* whenever target starts up */
2065 #define MGS_FIRST_OPC MGS_CONNECT
2067 #define MGS_PARAM_MAXLEN 1024
2068 #define KEY_SET_INFO "set_info"
2070 struct mgs_send_param {
2071 char mgs_param[MGS_PARAM_MAXLEN];
2074 /* We pass this info to the MGS so it can write config logs */
2075 #define MTI_NAME_MAXLEN 64
2076 #define MTI_PARAM_MAXLEN 4096
2077 #define MTI_NIDS_MAX 32
2078 struct mgs_target_info {
2079 __u32 mti_lustre_ver;
2080 __u32 mti_stripe_index;
2081 __u32 mti_config_ver;
2083 __u32 mti_nid_count;
2084 __u32 mti_instance; /* Running instance of target */
2085 char mti_fsname[MTI_NAME_MAXLEN];
2086 char mti_svname[MTI_NAME_MAXLEN];
2087 char mti_uuid[sizeof(struct obd_uuid)];
2088 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
2089 char mti_params[MTI_PARAM_MAXLEN];
2092 struct mgs_nidtbl_entry {
2093 __u64 mne_version; /* table version of this entry */
2094 __u32 mne_instance; /* target instance # */
2095 __u32 mne_index; /* target index */
2096 __u32 mne_length; /* length of this entry - by bytes */
2097 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2098 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2099 __u8 mne_nid_size; /* size of each NID, by bytes */
2100 __u8 mne_nid_count; /* # of NIDs in buffer */
2102 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2106 struct mgs_config_body {
2107 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2108 __u64 mcb_offset; /* next index of config log to request */
2109 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2111 __u8 mcb_bits; /* bits unit size of config log */
2112 __u32 mcb_units; /* # of units for bulk transfer */
2115 struct mgs_config_res {
2116 __u64 mcr_offset; /* index of last config log */
2117 __u64 mcr_size; /* size of the log */
2120 /* Config marker flags (in config log) */
2121 #define CM_START 0x01
2123 #define CM_SKIP 0x04
2124 #define CM_UPGRADE146 0x08
2125 #define CM_EXCLUDE 0x10
2126 #define CM_START_SKIP (CM_START | CM_SKIP)
2129 __u32 cm_step; /* aka config version */
2131 __u32 cm_vers; /* lustre release version number */
2132 __u32 cm_padding; /* 64 bit align */
2133 __s64 cm_createtime; /*when this record was first created */
2134 __s64 cm_canceltime; /*when this record is no longer valid*/
2135 char cm_tgtname[MTI_NAME_MAXLEN];
2136 char cm_comment[MTI_NAME_MAXLEN];
2140 * Opcodes for multiple servers.
2146 OBD_QC_CALLBACK, /* not used since 2.4 */
2150 #define OBD_FIRST_OPC OBD_PING
2153 * llog contexts indices.
2155 * There is compatibility problem with indexes below, they are not
2156 * continuous and must keep their numbers for compatibility needs.
2157 * See LU-5218 for details.
2160 LLOG_CONFIG_ORIG_CTXT = 0,
2161 LLOG_CONFIG_REPL_CTXT = 1,
2162 LLOG_MDS_OST_ORIG_CTXT = 2,
2163 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2164 LLOG_SIZE_ORIG_CTXT = 4,
2165 LLOG_SIZE_REPL_CTXT = 5,
2166 LLOG_TEST_ORIG_CTXT = 8,
2167 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2168 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2169 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2170 /* for multiple changelog consumers */
2171 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2172 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2176 /** Identifier for a single log object */
2178 struct ost_id lgl_oi;
2182 /** Records written to the CATALOGS list */
2183 #define CATLIST "CATALOGS"
2185 struct llog_logid lci_logid;
2191 /* Log data record types - there is no specific reason that these need to
2192 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2194 #define LLOG_OP_MAGIC 0x10600000
2195 #define LLOG_OP_MASK 0xfff00000
2198 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2199 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2200 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2201 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2202 REINT_UNLINK, /* obsolete after 2.5.0 */
2203 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2205 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2206 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2208 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2209 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2210 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2211 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2212 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2213 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2214 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2215 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2216 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2219 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2220 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2222 /** Log record header - stored in little endian order.
2223 * Each record must start with this struct, end with a llog_rec_tail,
2224 * and be a multiple of 256 bits in size.
2226 struct llog_rec_hdr {
2233 struct llog_rec_tail {
2238 /* Where data follow just after header */
2239 #define REC_DATA(ptr) \
2240 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2242 #define REC_DATA_LEN(rec) \
2243 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2244 sizeof(struct llog_rec_tail))
2246 struct llog_logid_rec {
2247 struct llog_rec_hdr lid_hdr;
2248 struct llog_logid lid_id;
2252 struct llog_rec_tail lid_tail;
2255 struct llog_unlink_rec {
2256 struct llog_rec_hdr lur_hdr;
2260 struct llog_rec_tail lur_tail;
2263 struct llog_unlink64_rec {
2264 struct llog_rec_hdr lur_hdr;
2265 struct lu_fid lur_fid;
2266 __u32 lur_count; /* to destroy the lost precreated */
2270 struct llog_rec_tail lur_tail;
2273 struct llog_setattr64_rec {
2274 struct llog_rec_hdr lsr_hdr;
2275 struct ost_id lsr_oi;
2281 struct llog_rec_tail lsr_tail;
2284 struct llog_size_change_rec {
2285 struct llog_rec_hdr lsc_hdr;
2286 struct ll_fid lsc_fid;
2291 struct llog_rec_tail lsc_tail;
2294 /* changelog llog name, needed by client replicators */
2295 #define CHANGELOG_CATALOG "changelog_catalog"
2297 struct changelog_setinfo {
2302 /** changelog record */
2303 struct llog_changelog_rec {
2304 struct llog_rec_hdr cr_hdr;
2305 struct changelog_rec cr; /**< Variable length field */
2306 struct llog_rec_tail cr_do_not_use; /**< for_sizezof_only */
2309 struct llog_changelog_user_rec {
2310 struct llog_rec_hdr cur_hdr;
2314 struct llog_rec_tail cur_tail;
2317 enum agent_req_status {
2325 static inline const char *agent_req_status2name(const enum agent_req_status ars)
2343 struct llog_agent_req_rec {
2344 struct llog_rec_hdr arr_hdr; /**< record header */
2345 __u32 arr_status; /**< status of the request */
2349 __u32 arr_archive_id; /**< backend archive number */
2350 __u64 arr_flags; /**< req flags */
2351 __u64 arr_compound_id;/**< compound cookie */
2352 __u64 arr_req_create; /**< req. creation time */
2353 __u64 arr_req_change; /**< req. status change time */
2354 struct hsm_action_item arr_hai; /**< req. to the agent */
2355 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2358 /* Old llog gen for compatibility */
2364 struct llog_gen_rec {
2365 struct llog_rec_hdr lgr_hdr;
2366 struct llog_gen lgr_gen;
2370 struct llog_rec_tail lgr_tail;
2373 /* flags for the logs */
2375 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2376 LLOG_F_IS_CAT = 0x2,
2377 LLOG_F_IS_PLAIN = 0x4,
2378 LLOG_F_EXT_JOBID = 0x8,
2379 LLOG_F_IS_FIXSIZE = 0x10,
2382 * Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
2383 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
2384 * because the catlog record is usually fixed size, but its plain
2385 * log record can be variable
2387 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID,
2390 /* On-disk header structure of each log object, stored in little endian order */
2391 #define LLOG_MIN_CHUNK_SIZE 8192
2392 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) +
2393 * sizeof(llh_tail) - sizeof(llh_bitmap)
2395 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
2396 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2398 /* flags for the logs */
2399 struct llog_log_hdr {
2400 struct llog_rec_hdr llh_hdr;
2401 __s64 llh_timestamp;
2403 __u32 llh_bitmap_offset;
2407 /* for a catalog the first plain slot is next to it */
2408 struct obd_uuid llh_tgtuuid;
2409 __u32 llh_reserved[LLOG_HEADER_SIZE / sizeof(__u32) - 23];
2410 /* These fields must always be at the end of the llog_log_hdr.
2411 * Note: llh_bitmap size is variable because llog chunk size could be
2412 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
2413 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
2414 * llh_tail should only be referred by LLOG_HDR_TAIL().
2415 * But this structure is also used by client/server llog interface
2416 * (see llog_client.c), it will be kept in its original way to avoid
2417 * compatibility issue.
2419 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
2420 struct llog_rec_tail llh_tail;
2423 #undef LLOG_HEADER_SIZE
2424 #undef LLOG_BITMAP_BYTES
2426 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2427 llh->llh_bitmap_offset - \
2428 sizeof(llh->llh_tail)) * 8)
2429 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
2430 (llh)->llh_bitmap_offset)
2431 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
2432 llh->llh_hdr.lrh_len - \
2433 sizeof(llh->llh_tail)))
2435 /** log cookies are used to reference a specific log file and a record
2438 struct llog_cookie {
2439 struct llog_logid lgc_lgl;
2445 /** llog protocol */
2446 enum llogd_rpc_ops {
2447 LLOG_ORIGIN_HANDLE_CREATE = 501,
2448 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2449 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2450 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2451 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2452 LLOG_ORIGIN_CONNECT = 506,
2453 LLOG_CATINFO = 507, /* deprecated */
2454 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2455 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2457 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2461 struct llog_logid lgd_logid;
2463 __u32 lgd_llh_flags;
2465 __u32 lgd_saved_index;
2467 __u64 lgd_cur_offset;
2470 struct llogd_conn_body {
2471 struct llog_gen lgdc_gen;
2472 struct llog_logid lgdc_logid;
2473 __u32 lgdc_ctxt_idx;
2476 /* Note: 64-bit types are 64-bit aligned in structure */
2478 __u64 o_valid; /* hot fields in this obdo */
2481 __u64 o_size; /* o_size-o_blocks == ost_lvb */
2485 __u64 o_blocks; /* brw: cli sent cached bytes */
2488 /* 32-bit fields start here: keep an even number of them via padding */
2489 __u32 o_blksize; /* optimal IO blocksize */
2490 __u32 o_mode; /* brw: cli sent cache remain */
2494 __u32 o_nlink; /* brw: checksum */
2496 __u32 o_misc; /* brw: o_dropped */
2498 __u64 o_ioepoch; /* epoch in ost writes */
2499 __u32 o_stripe_idx; /* holds stripe idx */
2501 struct lustre_handle o_handle; /* brw: lock handle to prolong locks
2503 struct llog_cookie o_lcookie; /* destroy: unlink cookie from MDS,
2504 * obsolete in 2.8, reused in OSP
2509 __u64 o_data_version; /* getattr: sum of iversion for
2511 * brw: grant space consumed on
2512 * the client for the write
2519 #define o_dirty o_blocks
2520 #define o_undirty o_mode
2521 #define o_dropped o_misc
2522 #define o_cksum o_nlink
2523 #define o_grant_used o_data_version
2525 /* request structure for OST's */
2530 /* Key for FIEMAP to be used in get_info calls */
2531 struct ll_fiemap_info_key {
2533 struct obdo lfik_oa;
2534 struct fiemap lfik_fiemap;
2537 /* security opcodes */
2540 SEC_CTX_INIT_CONT = 802,
2543 SEC_FIRST_OPC = SEC_CTX_INIT
2547 * capa related definitions
2549 #define CAPA_HMAC_MAX_LEN 64
2550 #define CAPA_HMAC_KEY_MAX_LEN 56
2552 /* NB take care when changing the sequence of elements this struct,
2553 * because the offset info is used in find_capa()
2555 struct lustre_capa {
2556 struct lu_fid lc_fid; /** fid */
2557 __u64 lc_opc; /** operations allowed */
2558 __u64 lc_uid; /** file owner */
2559 __u64 lc_gid; /** file group */
2560 __u32 lc_flags; /** HMAC algorithm & flags */
2561 __u32 lc_keyid; /** key# used for the capability */
2562 __u32 lc_timeout; /** capa timeout value (sec) */
2563 /* FIXME: y2038 time_t overflow: */
2564 __u32 lc_expiry; /** expiry time (sec) */
2565 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
2568 /** lustre_capa::lc_opc */
2570 CAPA_OPC_BODY_WRITE = 1 << 0, /**< write object data */
2571 CAPA_OPC_BODY_READ = 1 << 1, /**< read object data */
2572 CAPA_OPC_INDEX_LOOKUP = 1 << 2, /**< lookup object fid */
2573 CAPA_OPC_INDEX_INSERT = 1 << 3, /**< insert object fid */
2574 CAPA_OPC_INDEX_DELETE = 1 << 4, /**< delete object fid */
2575 CAPA_OPC_OSS_WRITE = 1 << 5, /**< write oss object data */
2576 CAPA_OPC_OSS_READ = 1 << 6, /**< read oss object data */
2577 CAPA_OPC_OSS_TRUNC = 1 << 7, /**< truncate oss object */
2578 CAPA_OPC_OSS_DESTROY = 1 << 8, /**< destroy oss object */
2579 CAPA_OPC_META_WRITE = 1 << 9, /**< write object meta data */
2580 CAPA_OPC_META_READ = 1 << 10, /**< read object meta data */
2583 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2584 #define CAPA_OPC_MDS_ONLY \
2585 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2586 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2587 #define CAPA_OPC_OSS_ONLY \
2588 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
2589 CAPA_OPC_OSS_DESTROY)
2590 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2591 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2593 struct lustre_capa_key {
2594 __u64 lk_seq; /**< mds# */
2595 __u32 lk_keyid; /**< key# */
2597 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
2600 /** The link ea holds 1 \a link_ea_entry for each hardlink */
2601 #define LINK_EA_MAGIC 0x11EAF1DFUL
2602 struct link_ea_header {
2605 __u64 leh_len; /* total size */
2606 __u32 leh_overflow_time;
2610 /** Hardlink data is name and parent fid.
2611 * Stored in this crazy struct for maximum packing and endian-neutrality
2613 struct link_ea_entry {
2614 /** __u16 stored big-endian, unaligned */
2615 unsigned char lee_reclen[2];
2616 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
2620 /** fid2path request/reply structure */
2621 struct getinfo_fid2path {
2622 struct lu_fid gf_fid;
2629 /** path2parent request/reply structures */
2631 struct lu_fid gp_fid; /**< parent FID */
2632 __u32 gp_linkno; /**< hardlink number */
2633 __u32 gp_name_size; /**< size of the name field */
2634 char gp_name[0]; /**< zero-terminated link name */
2638 LAYOUT_INTENT_ACCESS = 0,
2639 LAYOUT_INTENT_READ = 1,
2640 LAYOUT_INTENT_WRITE = 2,
2641 LAYOUT_INTENT_GLIMPSE = 3,
2642 LAYOUT_INTENT_TRUNC = 4,
2643 LAYOUT_INTENT_RELEASE = 5,
2644 LAYOUT_INTENT_RESTORE = 6
2647 /* enqueue layout lock with intent */
2648 struct layout_intent {
2649 __u32 li_opc; /* intent operation for enqueue, read, write etc */
2656 * On the wire version of hsm_progress structure.
2658 * Contains the userspace hsm_progress and some internal fields.
2660 struct hsm_progress_kernel {
2661 /* Field taken from struct hsm_progress */
2662 struct lu_fid hpk_fid;
2664 struct hsm_extent hpk_extent;
2666 __u16 hpk_errval; /* positive val */
2668 /* Additional fields */
2669 __u64 hpk_data_version;
2673 /** layout swap request structure
2674 * fid1 and fid2 are in mdt_body
2676 struct mdc_swap_layouts {
2681 struct lustre_handle cd_handle;
2682 struct lu_fid cd_fid;
2683 __u64 cd_data_version;
2684 __u64 cd_reserved[8];