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, 2016, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/lustre/lustre_idl.h
34 * Lustre wire protocol definitions.
37 /** \defgroup lustreidl lustreidl
39 * Lustre wire protocol definitions.
41 * ALL structs passing over the wire should be declared here. Structs
42 * that are used in interfaces with userspace should go in lustre_user.h.
44 * All structs being declared here should be built from simple fixed-size
45 * types (__u8, __u16, __u32, __u64) or be built from other types or
46 * structs also declared in this file. Similarly, all flags and magic
47 * values in those structs should also be declared here. This ensures
48 * that the Lustre wire protocol is not influenced by external dependencies.
50 * The only other acceptable items in this file are VERY SIMPLE accessor
51 * functions to avoid callers grubbing inside the structures. Nothing that
52 * depends on external functions or definitions should be in here.
54 * Structs must be properly aligned to put 64-bit values on an 8-byte
55 * boundary. Any structs being added here must also be added to
56 * utils/wirecheck.c and "make newwiretest" run to regenerate the
57 * utils/wiretest.c sources. This allows us to verify that wire structs
58 * have the proper alignment/size on all architectures.
60 * DO NOT CHANGE any of the structs, flags, values declared here and used
61 * in released Lustre versions. Some structs may have padding fields that
62 * can be used. Some structs might allow addition at the end (verify this
63 * in the code to ensure that new/old clients that see this larger struct
64 * do not fail, otherwise you need to implement protocol compatibility).
69 #ifndef _LUSTRE_IDL_H_
70 #define _LUSTRE_IDL_H_
72 #include <asm/byteorder.h>
73 #include <linux/types.h>
75 #include <libcfs/libcfs.h>
76 #include <lnet/types.h>
77 #include <lustre/lustre_user.h> /* Defn's shared with user-space. */
78 #include <lustre_ver.h>
83 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
84 * FOO_REPLY_PORTAL is for incoming replies on the FOO
85 * FOO_BULK_PORTAL is for incoming bulk on the FOO
88 #define CONNMGR_REQUEST_PORTAL 1
89 #define CONNMGR_REPLY_PORTAL 2
90 //#define OSC_REQUEST_PORTAL 3
91 #define OSC_REPLY_PORTAL 4
92 //#define OSC_BULK_PORTAL 5
93 #define OST_IO_PORTAL 6
94 #define OST_CREATE_PORTAL 7
95 #define OST_BULK_PORTAL 8
96 //#define MDC_REQUEST_PORTAL 9
97 #define MDC_REPLY_PORTAL 10
98 //#define MDC_BULK_PORTAL 11
99 #define MDS_REQUEST_PORTAL 12
100 //#define MDS_REPLY_PORTAL 13
101 #define MDS_BULK_PORTAL 14
102 #define LDLM_CB_REQUEST_PORTAL 15
103 #define LDLM_CB_REPLY_PORTAL 16
104 #define LDLM_CANCEL_REQUEST_PORTAL 17
105 #define LDLM_CANCEL_REPLY_PORTAL 18
106 //#define PTLBD_REQUEST_PORTAL 19
107 //#define PTLBD_REPLY_PORTAL 20
108 //#define PTLBD_BULK_PORTAL 21
109 #define MDS_SETATTR_PORTAL 22
110 #define MDS_READPAGE_PORTAL 23
111 #define OUT_PORTAL 24
112 #define MGC_REPLY_PORTAL 25
113 #define MGS_REQUEST_PORTAL 26
114 #define MGS_REPLY_PORTAL 27
115 #define OST_REQUEST_PORTAL 28
116 #define FLD_REQUEST_PORTAL 29
117 #define SEQ_METADATA_PORTAL 30
118 #define SEQ_DATA_PORTAL 31
119 #define SEQ_CONTROLLER_PORTAL 32
120 #define MGS_BULK_PORTAL 33
122 /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com, n8851@cray.com */
125 #define PTL_RPC_MSG_REQUEST 4711
126 #define PTL_RPC_MSG_ERR 4712
127 #define PTL_RPC_MSG_REPLY 4713
129 /* DON'T use swabbed values of MAGIC as magic! */
130 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
131 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
133 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
135 #define PTLRPC_MSG_VERSION 0x00000003
136 #define LUSTRE_VERSION_MASK 0xffff0000
137 #define LUSTRE_OBD_VERSION 0x00010000
138 #define LUSTRE_MDS_VERSION 0x00020000
139 #define LUSTRE_OST_VERSION 0x00030000
140 #define LUSTRE_DLM_VERSION 0x00040000
141 #define LUSTRE_LOG_VERSION 0x00050000
142 #define LUSTRE_MGS_VERSION 0x00060000
145 * Describes a range of sequence, lsr_start is included but lsr_end is
147 * Same structure is used in fld module where lsr_index field holds mdt id
150 struct lu_seq_range {
157 struct lu_seq_range_array {
160 struct lu_seq_range lsra_lsr[0];
163 #define LU_SEQ_RANGE_MDT 0x0
164 #define LU_SEQ_RANGE_OST 0x1
165 #define LU_SEQ_RANGE_ANY 0x3
167 #define LU_SEQ_RANGE_MASK 0x3
169 /** \defgroup lu_fid lu_fid
173 * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat.
174 * Deprecated since HSM and SOM attributes are now stored in separate on-disk
178 LMAC_HSM = 0x00000001,
179 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
180 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
181 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
182 * under /O/<seq>/d<x>. */
186 * Masks for all features that should be supported by a Lustre version to
187 * access a specific file.
188 * This information is stored in lustre_mdt_attrs::lma_incompat.
191 LMAI_RELEASED = 0x00000001, /* file is released */
192 LMAI_AGENT = 0x00000002, /* agent inode */
193 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
194 is on the remote MDT */
195 LMAI_STRIPED = 0x00000008, /* striped directory inode */
196 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
197 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
198 LMAI_STRIPED | LMAI_ORPHAN)
201 extern void lustre_lma_swab(struct lustre_mdt_attrs *lma);
202 extern void lustre_lma_init(struct lustre_mdt_attrs *lma,
203 const struct lu_fid *fid,
204 __u32 compat, __u32 incompat);
206 /* copytool uses a 32b bitmask field to encode archive-Ids during register
208 * archive num = 0 => all
209 * archive num from 1 to 32
211 #define LL_HSM_MAX_ARCHIVE (sizeof(__u32) * 8)
214 * HSM on-disk attributes stored in a separate xattr.
217 /** Bitfield for supported data in this structure. For future use. */
220 /** HSM flags, see hsm_flags enum below */
222 /** backend archive id associated with the file */
224 /** version associated with the last archiving, if any */
227 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
233 /** LASTID file has zero OID */
234 LUSTRE_FID_LASTID_OID = 0UL,
235 /** initial fid id value */
236 LUSTRE_FID_INIT_OID = 1UL
240 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
241 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
242 * used for other purposes and not risk collisions with existing inodes.
244 * Different FID Format
245 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
248 FID_SEQ_OST_MDT0 = 0,
249 FID_SEQ_LLOG = 1, /* unnamed llogs */
251 FID_SEQ_UNUSED_START = 3,
252 FID_SEQ_UNUSED_END = 9,
253 FID_SEQ_LLOG_NAME = 10, /* named llogs */
256 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
257 FID_SEQ_IDIF = 0x100000000ULL,
258 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
259 /* Normal FID sequence starts from this value, i.e. 1<<33 */
260 FID_SEQ_START = 0x200000000ULL,
261 /* sequence for local pre-defined FIDs listed in local_oid */
262 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
263 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
264 /* sequence is used for local named objects FIDs generated
265 * by local_object_storage library */
266 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
267 /* Because current FLD will only cache the fid sequence, instead
268 * of oid on the client side, if the FID needs to be exposed to
269 * clients sides, it needs to make sure all of fids under one
270 * sequence will be located in one MDT. */
271 FID_SEQ_SPECIAL = 0x200000004ULL,
272 FID_SEQ_QUOTA = 0x200000005ULL,
273 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
274 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
275 FID_SEQ_LAYOUT_RBTREE = 0x200000008ULL,
276 /* sequence is used for update logs of cross-MDT operation */
277 FID_SEQ_UPDATE_LOG = 0x200000009ULL,
278 /* Sequence is used for the directory under which update logs
280 FID_SEQ_UPDATE_LOG_DIR = 0x20000000aULL,
281 FID_SEQ_NORMAL = 0x200000400ULL,
282 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
285 #define OBIF_OID_MAX_BITS 32
286 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
287 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
288 #define IDIF_OID_MAX_BITS 48
289 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
290 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
292 /** OID for FID_SEQ_SPECIAL */
294 /* Big Filesystem Lock to serialize rename operations */
295 FID_OID_SPECIAL_BFL = 1UL,
298 /** OID for FID_SEQ_DOT_LUSTRE */
299 enum dot_lustre_oid {
300 FID_OID_DOT_LUSTRE = 1UL,
301 FID_OID_DOT_LUSTRE_OBF = 2UL,
302 FID_OID_DOT_LUSTRE_LPF = 3UL,
305 /** OID for FID_SEQ_ROOT */
308 FID_OID_ECHO_ROOT = 2UL,
311 struct lu_orphan_rec {
312 /* The MDT-object's FID referenced by the orphan OST-object */
313 struct lu_fid lor_fid;
318 struct lu_orphan_ent {
319 /* The orphan OST-object's FID */
320 struct lu_fid loe_key;
321 struct lu_orphan_rec loe_rec;
326 /** \defgroup lu_dir lu_dir
330 * Enumeration of possible directory entry attributes.
332 * Attributes follow directory entry header in the order they appear in this
335 enum lu_dirent_attrs {
338 LUDA_64BITHASH = 0x0004,
340 /* The following attrs are used for MDT internal only,
341 * not visible to client */
343 /* Something in the record is unknown, to be verified in further. */
344 LUDA_UNKNOWN = 0x0400,
345 /* Ignore this record, go to next directly. */
346 LUDA_IGNORE = 0x0800,
347 /* The system is upgraded, has beed or to be repaired (dryrun). */
348 LUDA_UPGRADE = 0x1000,
349 /* The dirent has been repaired, or to be repaired (dryrun). */
350 LUDA_REPAIR = 0x2000,
351 /* Only check but not repair the dirent inconsistency */
352 LUDA_VERIFY_DRYRUN = 0x4000,
353 /* Verify the dirent consistency */
354 LUDA_VERIFY = 0x8000,
357 #define LU_DIRENT_ATTRS_MASK 0xff00
360 * Layout of readdir pages, as transmitted on wire.
363 /** valid if LUDA_FID is set. */
364 struct lu_fid lde_fid;
365 /** a unique entry identifier: a hash or an offset. */
367 /** total record length, including all attributes. */
371 /** optional variable size attributes following this entry.
372 * taken from enum lu_dirent_attrs.
375 /** name is followed by the attributes indicated in ->ldp_attrs, in
376 * their natural order. After the last attribute, padding bytes are
377 * added to make ->lde_reclen a multiple of 8.
383 * Definitions of optional directory entry attributes formats.
385 * Individual attributes do not have their length encoded in a generic way. It
386 * is assumed that consumer of an attribute knows its format. This means that
387 * it is impossible to skip over an unknown attribute, except by skipping over all
388 * remaining attributes (by using ->lde_reclen), which is not too
389 * constraining, because new server versions will append new attributes at
390 * the end of an entry.
394 * Fid directory attribute: a fid of an object referenced by the entry. This
395 * will be almost always requested by the client and supplied by the server.
397 * Aligned to 8 bytes.
399 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
404 * Aligned to 2 bytes.
411 __u64 ldp_hash_start;
415 struct lu_dirent ldp_entries[0];
418 enum lu_dirpage_flags {
420 * dirpage contains no entry.
424 * last entry's lde_hash equals ldp_hash_end.
429 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
431 if (__le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
434 return dp->ldp_entries;
437 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
439 struct lu_dirent *next;
441 if (__le16_to_cpu(ent->lde_reclen) != 0)
442 next = ((void *)ent) + __le16_to_cpu(ent->lde_reclen);
449 static inline size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
453 if (attr & LUDA_TYPE) {
454 const size_t align = sizeof(struct luda_type) - 1;
455 size = (sizeof(struct lu_dirent) + namelen + align) & ~align;
456 size += sizeof(struct luda_type);
458 size = sizeof(struct lu_dirent) + namelen;
460 return (size + 7) & ~7;
463 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
466 * MDS_READPAGE page size
468 * This is the directory page size packed in MDS_READPAGE RPC.
469 * It's different than PAGE_SIZE because the client needs to
470 * access the struct lu_dirpage header packed at the beginning of
471 * the "page" and without this there isn't any way to know find the
472 * lu_dirpage header is if client and server PAGE_SIZE differ.
474 #define LU_PAGE_SHIFT 12
475 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
476 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
478 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
482 struct lustre_handle {
485 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
487 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
489 return lh->cookie != 0;
492 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
493 const struct lustre_handle *lh2)
495 return lh1->cookie == lh2->cookie;
498 static inline void lustre_handle_copy(struct lustre_handle *tgt,
499 const struct lustre_handle *src)
501 tgt->cookie = src->cookie;
504 struct lustre_handle_array {
506 struct lustre_handle handles[0];
509 /* flags for lm_flags */
510 #define MSGHDR_AT_SUPPORT 0x1
511 #define MSGHDR_CKSUM_INCOMPAT18 0x2
513 #define lustre_msg lustre_msg_v2
514 /* we depend on this structure to be 8-byte aligned */
515 /* this type is only endian-adjusted in lustre_unpack_msg() */
516 struct lustre_msg_v2 {
528 /* without gss, ptlrpc_body is put at the first buffer. */
529 #define PTLRPC_NUM_VERSIONS 4
530 struct ptlrpc_body_v3 {
531 struct lustre_handle pb_handle;
536 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
537 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
540 __u64 pb_last_committed;
545 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
546 __u32 pb_service_time; /* for rep, actual service time */
549 /* VBR: pre-versions */
550 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
551 __u64 pb_mbits; /**< match bits for bulk request */
552 /* padding for future needs */
553 __u64 pb_padding64_0;
554 __u64 pb_padding64_1;
555 __u64 pb_padding64_2;
556 char pb_jobid[LUSTRE_JOBID_SIZE];
558 #define ptlrpc_body ptlrpc_body_v3
560 struct ptlrpc_body_v2 {
561 struct lustre_handle pb_handle;
566 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
567 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
570 __u64 pb_last_committed;
575 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
576 __u32 pb_service_time; /* for rep, actual service time, also used for
577 net_latency of req */
580 /* VBR: pre-versions */
581 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
582 __u64 pb_mbits; /**< unused in V2 */
583 /* padding for future needs */
584 __u64 pb_padding64_0;
585 __u64 pb_padding64_1;
586 __u64 pb_padding64_2;
589 /* message body offset for lustre_msg_v2 */
590 /* ptlrpc body offset in all request/reply messages */
591 #define MSG_PTLRPC_BODY_OFF 0
593 /* normal request/reply message record offset */
594 #define REQ_REC_OFF 1
595 #define REPLY_REC_OFF 1
597 /* ldlm request message body offset */
598 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
599 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
601 /* ldlm intent lock message body offset */
602 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
603 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
605 /* ldlm reply message body offset */
606 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
607 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
609 /** only use in req->rq_{req,rep}_swab_mask */
610 #define MSG_PTLRPC_HEADER_OFF 31
612 /* Flags that are operation-specific go in the top 16 bits. */
613 #define MSG_OP_FLAG_MASK 0xffff0000
614 #define MSG_OP_FLAG_SHIFT 16
616 /* Flags that apply to all requests are in the bottom 16 bits */
617 #define MSG_GEN_FLAG_MASK 0x0000ffff
618 #define MSG_LAST_REPLAY 0x0001
619 #define MSG_RESENT 0x0002
620 #define MSG_REPLAY 0x0004
621 /* #define MSG_AT_SUPPORT 0x0008
622 * This was used in early prototypes of adaptive timeouts, and while there
623 * shouldn't be any users of that code there also isn't a need for using this
624 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
625 #define MSG_DELAY_REPLAY 0x0010
626 #define MSG_VERSION_REPLAY 0x0020
627 #define MSG_REQ_REPLAY_DONE 0x0040
628 #define MSG_LOCK_REPLAY_DONE 0x0080
631 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
634 #define MSG_CONNECT_RECOVERING 0x00000001
635 #define MSG_CONNECT_RECONNECT 0x00000002
636 #define MSG_CONNECT_REPLAYABLE 0x00000004
637 //#define MSG_CONNECT_PEER 0x8
638 #define MSG_CONNECT_LIBCLIENT 0x00000010
639 #define MSG_CONNECT_INITIAL 0x00000020
640 #define MSG_CONNECT_ASYNC 0x00000040
641 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
642 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
645 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
646 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
647 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
648 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
649 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
650 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
651 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
652 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
653 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
654 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
655 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
656 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
657 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
658 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
659 *We do not support JOIN FILE
660 *anymore, reserve this flags
661 *just for preventing such bit
663 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
664 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
665 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
666 * in production. Removed in
667 * 2.9. Keep this flag to
670 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
671 * never used in production.
672 * Removed in 2.9. Keep this
673 * flag to avoid reusing.
675 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
676 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
677 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
678 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
679 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
680 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
681 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
682 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
683 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
684 #define OBD_CONNECT_REAL 0x8000000ULL /* obsolete since 2.8 */
685 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
686 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
687 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
688 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
689 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
690 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
691 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
692 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
693 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
694 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
695 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
697 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
698 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
699 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
700 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
701 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
702 * RPC error properly */
703 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
704 * finer space reservation */
705 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
706 * policy and 2.x server */
707 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
708 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
709 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
710 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
711 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
712 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* improved flock deadlock detection */
713 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/* create stripe disposition*/
714 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
716 #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */
717 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */
718 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify
720 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
721 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
722 #define OBD_CONNECT_LOCK_AHEAD 0x1000000000000000ULL /* lock ahead */
723 /** bulk matchbits is sent within ptlrpc_body */
724 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL
725 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
726 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
727 /* ocd_connect_flags2 flags */
728 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* set file security context at create */
731 * Please DO NOT add flag values here before first ensuring that this same
732 * flag value is not in use on some other branch. Please clear any such
733 * changes with senior engineers before starting to use a new flag. Then,
734 * submit a small patch against EVERY branch that ONLY adds the new flag,
735 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
736 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
737 * can be approved and landed easily to reserve the flag for future use. */
739 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
740 * connection. It is a temporary bug fix for Imperative Recovery interop
741 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
742 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
743 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
745 #define OCD_HAS_FLAG(ocd, flg) \
746 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
749 #ifdef HAVE_LRU_RESIZE_SUPPORT
750 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
752 #define LRU_RESIZE_CONNECT_FLAG 0
755 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
756 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
757 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
758 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
759 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
760 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
761 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
762 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
763 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
764 OBD_CONNECT_EINPROGRESS | \
765 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
766 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
767 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
768 OBD_CONNECT_FLOCK_DEAD | \
769 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
770 OBD_CONNECT_OPEN_BY_FID | \
771 OBD_CONNECT_DIR_STRIPE | \
772 OBD_CONNECT_BULK_MBITS | \
773 OBD_CONNECT_MULTIMODRPCS | \
774 OBD_CONNECT_SUBTREE | \
777 #define MDT_CONNECT_SUPPORTED2 OBD_CONNECT2_FILE_SECCTX
779 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
780 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
781 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
782 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
783 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
784 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
785 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
786 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
787 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
788 OBD_CONNECT_MAX_EASIZE | \
789 OBD_CONNECT_EINPROGRESS | \
790 OBD_CONNECT_JOBSTATS | \
791 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
792 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
793 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
794 OBD_CONNECT_BULK_MBITS | \
795 OBD_CONNECT_GRANT_PARAM)
796 #define OST_CONNECT_SUPPORTED2 0
798 #define ECHO_CONNECT_SUPPORTED 0
799 #define ECHO_CONNECT_SUPPORTED2 0
801 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
802 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
803 OBD_CONNECT_MNE_SWAB | OBD_CONNECT_PINGLESS |\
804 OBD_CONNECT_BULK_MBITS)
806 #define MGS_CONNECT_SUPPORTED2 0
808 /* Features required for this version of the client to work with server */
809 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
812 /* This structure is used for both request and reply.
814 * If we eventually have separate connect data for different types, which we
815 * almost certainly will, then perhaps we stick a union in here. */
816 struct obd_connect_data {
817 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
818 __u32 ocd_version; /* lustre release version number */
819 __u32 ocd_grant; /* initial cache grant amount (bytes) */
820 __u32 ocd_index; /* LOV index to connect to */
821 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
822 __u64 ocd_ibits_known; /* inode bits this client understands */
823 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
824 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
825 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
826 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
827 __u64 ocd_transno; /* first transno from client to be replayed */
828 __u32 ocd_group; /* MDS group on OST */
829 __u32 ocd_cksum_types; /* supported checksum algorithms */
830 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
831 __u32 ocd_instance; /* instance # of this target */
832 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
833 /* Fields after ocd_maxbytes are only accessible by the receiver
834 * if the corresponding flag in ocd_connect_flags is set. Accessing
835 * any field after ocd_maxbytes on the receiver without a valid flag
836 * may result in out-of-bound memory access and kernel oops. */
837 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
838 __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */
839 __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */
840 __u64 ocd_connect_flags2;
841 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
842 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
843 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
844 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
845 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
846 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
847 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
848 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
849 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
850 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
851 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
852 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
853 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
856 * Please DO NOT use any fields here before first ensuring that this same
857 * field is not in use on some other branch. Please clear any such changes
858 * with senior engineers before starting to use a new field. Then, submit
859 * a small patch against EVERY branch that ONLY adds the new field along with
860 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
861 * reserve the flag for future use. */
864 * Supported checksum algorithms. Up to 32 checksum types are supported.
865 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
866 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
867 * algorithm and also the OBD_FL_CKSUM* flags.
870 OBD_CKSUM_CRC32 = 0x00000001,
871 OBD_CKSUM_ADLER = 0x00000002,
872 OBD_CKSUM_CRC32C= 0x00000004,
876 * OST requests: OBDO & OBD request records
881 OST_REPLY = 0, /* reply ? */
897 OST_QUOTACHECK = 18, /* not used since 2.4 */
899 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
901 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
903 #define OST_FIRST_OPC OST_REPLY
906 OBD_FL_INLINEDATA = 0x00000001,
907 OBD_FL_OBDMDEXISTS = 0x00000002,
908 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
909 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
910 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
911 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
912 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
913 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
914 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
915 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
916 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
917 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
918 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
919 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
920 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
921 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
922 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
923 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
924 * XXX: obsoleted - reserved for old
925 * clients prior than 2.2 */
926 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
927 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
928 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
929 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
931 /* Note that while these checksum values are currently separate bits,
932 * in 2.x we can actually allow all values from 1-31 if we wanted. */
933 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
936 /* mask for local-only flag, which won't be sent over network */
937 OBD_FL_LOCAL_MASK = 0xF0000000,
941 * All LOV EA magics should have the same postfix, if some new version
942 * Lustre instroduces new LOV EA magic, then when down-grade to an old
943 * Lustre, even though the old version system does not recognizes such
944 * new magic, it still can distinguish the corrupted cases by checking
945 * the magic's postfix.
947 #define LOV_MAGIC_MAGIC 0x0BD0
948 #define LOV_MAGIC_MASK 0xFFFF
950 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
951 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
952 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
953 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
954 /* reserved for specifying OSTs */
955 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
956 #define LOV_MAGIC LOV_MAGIC_V1
959 * magic for fully defined striping
960 * the idea is that we should have different magics for striping "hints"
961 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
962 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
963 * we can't just change it w/o long way preparation, but we still need a
964 * mechanism to allow LOD to differentiate hint versus ready striping.
965 * so, at the moment we do a trick: MDT knows what to expect from request
966 * depending on the case (replay uses ready striping, non-replay req uses
967 * hints), so MDT replaces magic with appropriate one and now LOD can
968 * easily understand what's inside -bzzz
970 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
971 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
973 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
974 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
976 #define lov_ost_data lov_ost_data_v1
977 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
978 struct ost_id l_ost_oi; /* OST object ID */
979 __u32 l_ost_gen; /* generation of this l_ost_idx */
980 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
983 #define lov_mds_md lov_mds_md_v1
984 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
985 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
986 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
987 struct ost_id lmm_oi; /* LOV object ID */
988 __u32 lmm_stripe_size; /* size of stripe in bytes */
989 /* lmm_stripe_count used to be __u32 */
990 __u16 lmm_stripe_count; /* num stripes in use for this object */
991 __u16 lmm_layout_gen; /* layout generation number */
992 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
995 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
996 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
998 /* This is the default MDT reply size allocated, should the striping be bigger,
999 * it will be reallocated in mdt_fix_reply.
1000 * 100 stripes is a bit less than 2.5k of data */
1001 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1002 100 * sizeof(struct lov_ost_data))
1004 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1005 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1006 #define XATTR_USER_PREFIX "user."
1007 #define XATTR_TRUSTED_PREFIX "trusted."
1008 #define XATTR_SECURITY_PREFIX "security."
1010 #define XATTR_NAME_LOV "trusted.lov"
1011 #define XATTR_NAME_LMA "trusted.lma"
1012 #define XATTR_NAME_LMV "trusted.lmv"
1013 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1014 #define XATTR_NAME_LINK "trusted.link"
1015 #define XATTR_NAME_FID "trusted.fid"
1016 #define XATTR_NAME_VERSION "trusted.version"
1017 #define XATTR_NAME_SOM "trusted.som"
1018 #define XATTR_NAME_HSM "trusted.hsm"
1019 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1020 #define XATTR_NAME_DUMMY "trusted.dummy"
1022 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1023 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1025 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1026 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1027 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1028 struct ost_id lmm_oi; /* LOV object ID */
1029 __u32 lmm_stripe_size; /* size of stripe in bytes */
1030 /* lmm_stripe_count used to be __u32 */
1031 __u16 lmm_stripe_count; /* num stripes in use for this object */
1032 __u16 lmm_layout_gen; /* layout generation number */
1033 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1034 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1037 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1039 if (lmm_magic == LOV_MAGIC_V3)
1040 return sizeof(struct lov_mds_md_v3) +
1041 stripes * sizeof(struct lov_ost_data_v1);
1043 return sizeof(struct lov_mds_md_v1) +
1044 stripes * sizeof(struct lov_ost_data_v1);
1048 lov_mds_md_max_stripe_count(size_t buf_size, __u32 lmm_magic)
1050 switch (lmm_magic) {
1051 case LOV_MAGIC_V1: {
1052 struct lov_mds_md_v1 lmm;
1054 if (buf_size < sizeof(lmm))
1057 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1059 case LOV_MAGIC_V3: {
1060 struct lov_mds_md_v3 lmm;
1062 if (buf_size < sizeof(lmm))
1065 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1072 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1073 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1074 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1075 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1076 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1077 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1078 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1079 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1080 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1081 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1082 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1083 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1084 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1085 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1086 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
1087 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1088 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1089 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1090 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1091 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1092 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1093 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1094 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1095 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1096 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1097 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1098 /* ->mds if epoch opens or closes */
1099 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1100 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1101 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1102 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1103 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1105 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1106 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1107 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1108 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1110 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1111 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1112 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1113 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1114 /* OBD_MD_FLRMTPERM (0x0000010000000000ULL) remote perm, obsolete */
1115 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1116 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1117 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1118 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1119 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1120 * under lock; for xattr
1121 * requests means the
1122 * client holds the lock */
1123 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1125 /* OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) lfs lsetfacl, obsolete */
1126 /* OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) lfs lgetfacl, obsolete */
1127 /* OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) lfs rsetfacl, obsolete */
1128 /* OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) lfs rgetfacl, obsolete */
1130 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1131 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1134 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1136 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1137 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1138 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1139 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1140 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1142 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1144 /* don't forget obdo_fid which is way down at the bottom so it can
1145 * come after the definition of llog_cookie */
1149 HSS_CLEARMASK = 0x02,
1150 HSS_ARCHIVE_ID = 0x04,
1153 struct hsm_state_set {
1155 __u32 hss_archive_id;
1157 __u64 hss_clearmask;
1160 /* ost_body.data values for OST_BRW */
1162 #define OBD_BRW_READ 0x01
1163 #define OBD_BRW_WRITE 0x02
1164 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1165 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1166 * transfer and is not accounted in
1168 #define OBD_BRW_CHECK 0x10
1169 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1170 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1171 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1172 #define OBD_BRW_NOQUOTA 0x100
1173 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1174 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1175 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1176 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1177 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1178 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1179 * that the client is running low on
1180 * space for unstable pages; asking
1181 * it to sync quickly */
1183 #define OBD_OBJECT_EOF LUSTRE_EOF
1185 #define OST_MIN_PRECREATE 32
1186 #define OST_MAX_PRECREATE 20000
1189 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1190 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1191 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1192 * high 16 bits in 2.4 and later */
1193 __u32 ioo_bufcnt; /* number of niobufs for this object */
1196 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1197 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1198 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. */
1199 #define IOOBJ_MAX_BRW_BITS 16
1200 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1201 #define ioobj_max_brw_set(ioo, num) \
1202 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1204 /* multiple of 8 bytes => can array */
1205 struct niobuf_remote {
1211 /* lock value block communicated between the filter and llite */
1213 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1214 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1215 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1216 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1217 #define OST_LVB_IS_ERR(blocks) \
1218 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1219 #define OST_LVB_SET_ERR(blocks, rc) \
1220 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1221 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1244 * lquota data structures
1247 #ifndef QUOTABLOCK_BITS
1248 # define QUOTABLOCK_BITS LUSTRE_QUOTABLOCK_BITS
1251 #ifndef QUOTABLOCK_SIZE
1252 # define QUOTABLOCK_SIZE LUSTRE_QUOTABLOCK_SIZE
1256 # define toqb lustre_stoqb
1259 /* The lquota_id structure is an union of all the possible identifier types that
1260 * can be used with quota, this includes:
1263 * - a FID which can be used for per-directory quota in the future */
1265 struct lu_fid qid_fid; /* FID for per-directory quota */
1266 __u64 qid_uid; /* user identifier */
1267 __u64 qid_gid; /* group identifier */
1270 /* quotactl management */
1271 struct obd_quotactl {
1273 __u32 qc_type; /* see Q_* flag below */
1276 struct obd_dqinfo qc_dqinfo;
1277 struct obd_dqblk qc_dqblk;
1280 #define Q_COPY(out, in, member) (out)->member = (in)->member
1282 #define QCTL_COPY(out, in) \
1284 Q_COPY(out, in, qc_cmd); \
1285 Q_COPY(out, in, qc_type); \
1286 Q_COPY(out, in, qc_id); \
1287 Q_COPY(out, in, qc_stat); \
1288 Q_COPY(out, in, qc_dqinfo); \
1289 Q_COPY(out, in, qc_dqblk); \
1292 /* Body of quota request used for quota acquire/release RPCs between quota
1293 * master (aka QMT) and slaves (ak QSD). */
1295 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1296 * and type (data or metadata) as well as
1297 * the quota type (user or group). */
1298 union lquota_id qb_id; /* uid or gid or directory FID */
1299 __u32 qb_flags; /* see below */
1301 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1302 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1303 __u64 qb_slv_ver; /* slave index file version */
1304 struct lustre_handle qb_lockh; /* per-ID lock handle */
1305 struct lustre_handle qb_glb_lockh; /* global lock handle */
1306 __u64 qb_padding1[4];
1309 /* When the quota_body is used in the reply of quota global intent
1310 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1311 #define qb_slv_fid qb_fid
1312 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1314 #define qb_qunit qb_usage
1316 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1317 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1318 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1319 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1321 /* Quota types currently supported */
1323 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1324 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1328 /* There are 2 different resource types on which a quota limit can be enforced:
1329 * - inodes on the MDTs
1330 * - blocks on the OSTs */
1332 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1333 LQUOTA_RES_DT = 0x02,
1335 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1337 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1340 * Space accounting support
1341 * Format of an accounting record, providing disk usage information for a given
1344 struct lquota_acct_rec { /* 16 bytes */
1345 __u64 bspace; /* current space in use */
1346 __u64 ispace; /* current # inodes in use */
1350 * Global quota index support
1351 * Format of a global record, providing global quota settings for a given quota
1354 struct lquota_glb_rec { /* 32 bytes */
1355 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1356 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1357 __u64 qbr_time; /* grace time, in seconds */
1358 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1363 * Slave index support
1364 * Format of a slave record, recording how much space is granted to a given
1367 struct lquota_slv_rec { /* 8 bytes */
1368 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1369 * in #inodes or kbytes */
1372 /* Data structures associated with the quota locks */
1374 /* Glimpse descriptor used for the index & per-ID quota locks */
1375 struct ldlm_gl_lquota_desc {
1376 union lquota_id gl_id; /* quota ID subject to the glimpse */
1377 __u64 gl_flags; /* see LQUOTA_FL* below */
1378 __u64 gl_ver; /* new index version */
1379 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1380 __u64 gl_softlimit; /* new softlimit */
1384 #define gl_qunit gl_hardlimit /* current qunit value used when
1385 * glimpsing per-ID quota locks */
1387 /* quota glimpse flags */
1388 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1390 /* LVB used with quota (global and per-ID) locks */
1392 __u64 lvb_flags; /* see LQUOTA_FL* above */
1393 __u64 lvb_id_may_rel; /* space that might be released later */
1394 __u64 lvb_id_rel; /* space released by the slave for this ID */
1395 __u64 lvb_id_qunit; /* current qunit value */
1399 /* LVB used with global quota lock */
1400 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1408 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1417 MDS_GETATTR_NAME = 34,
1422 MDS_DISCONNECT = 39,
1425 MDS_PIN = 42, /* obsolete, never used in a release */
1426 MDS_UNPIN = 43, /* obsolete, never used in a release */
1428 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1430 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1433 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1435 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1437 MDS_HSM_STATE_GET = 54,
1438 MDS_HSM_STATE_SET = 55,
1439 MDS_HSM_ACTION = 56,
1440 MDS_HSM_PROGRESS = 57,
1441 MDS_HSM_REQUEST = 58,
1442 MDS_HSM_CT_REGISTER = 59,
1443 MDS_HSM_CT_UNREGISTER = 60,
1444 MDS_SWAP_LAYOUTS = 61,
1448 #define MDS_FIRST_OPC MDS_GETATTR
1451 /* opcodes for object update */
1457 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1474 } mds_reint_t, mdt_reint_t;
1476 /* the disposition of the intent outlines what was executed */
1477 #define DISP_IT_EXECD 0x00000001
1478 #define DISP_LOOKUP_EXECD 0x00000002
1479 #define DISP_LOOKUP_NEG 0x00000004
1480 #define DISP_LOOKUP_POS 0x00000008
1481 #define DISP_OPEN_CREATE 0x00000010
1482 #define DISP_OPEN_OPEN 0x00000020
1483 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1484 #define DISP_ENQ_OPEN_REF 0x00800000
1485 #define DISP_ENQ_CREATE_REF 0x01000000
1486 #define DISP_OPEN_LOCK 0x02000000
1487 #define DISP_OPEN_LEASE 0x04000000
1488 #define DISP_OPEN_STRIPE 0x08000000
1489 #define DISP_OPEN_DENY 0x10000000
1491 /* INODE LOCK PARTS */
1492 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
1493 * was used to protect permission (mode,
1494 * owner, group etc) before 2.4. */
1495 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1496 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1497 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1499 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
1500 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
1501 * Because for remote directories(in DNE), these locks will be granted by
1502 * different MDTs(different ldlm namespace).
1504 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
1505 * For Remote directory, the master MDT, where the remote directory is, will
1506 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
1507 * will grant LOOKUP_LOCK. */
1508 #define MDS_INODELOCK_PERM 0x000010
1509 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
1511 #define MDS_INODELOCK_MAXSHIFT 5
1512 /* This FULL lock is useful to take on unlink sort of operations */
1513 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1515 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1516 * but was moved into name[1] along with the OID to avoid consuming the
1517 * name[2,3] fields that need to be used for the quota id (also a FID). */
1519 LUSTRE_RES_ID_SEQ_OFF = 0,
1520 LUSTRE_RES_ID_VER_OID_OFF = 1,
1521 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1522 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1523 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1524 LUSTRE_RES_ID_HSH_OFF = 3
1527 #define MDS_STATUS_CONN 1
1528 #define MDS_STATUS_LOV 2
1531 /* these should be identical to their EXT4_*_FL counterparts, they are
1532 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1533 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1534 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1535 LUSTRE_APPEND_FL = 0x00000020, /* writes to file may only append */
1536 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1537 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1538 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1539 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1540 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1541 LUSTRE_DIRECTIO_FL = 0x00100000, /* Use direct i/o */
1542 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1544 /* These flags will not be identical to any EXT4_*_FL counterparts,
1545 * and only reserved for lustre purpose. Note: these flags might
1546 * be conflict with some of EXT4 flags, so
1547 * 1. these conflict flags needs to be removed when the flag is
1548 * wired by la_flags see osd_attr_get().
1549 * 2. If these flags needs to be stored into inode, they will be
1550 * stored in LMA. see LMAI_XXXX */
1551 LUSTRE_ORPHAN_FL = 0x00002000,
1553 LUSTRE_LMA_FL_MASKS = LUSTRE_ORPHAN_FL,
1557 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1558 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1559 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1560 * the S_* flags are kernel-internal values that change between kernel
1561 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1562 * See b=16526 for a full history. */
1563 static inline int ll_ext_to_inode_flags(int flags)
1565 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1566 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1567 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1568 #if defined(S_DIRSYNC)
1569 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1571 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0));
1574 static inline int ll_inode_to_ext_flags(int iflags)
1576 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1577 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1578 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1579 #if defined(S_DIRSYNC)
1580 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1582 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0));
1586 /* 64 possible states */
1587 enum md_transient_state {
1588 MS_RESTORE = (1 << 0), /* restore is running */
1592 struct lu_fid mbo_fid1;
1593 struct lu_fid mbo_fid2;
1594 struct lustre_handle mbo_handle;
1596 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1600 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1602 __u64 mbo_t_state; /* transient file state defined in
1603 * enum md_transient_state
1604 * was "ino" until 2.4.0 */
1607 __u32 mbo_capability;
1611 __u32 mbo_flags; /* LUSTRE_*_FL file attributes */
1613 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1614 __u32 mbo_unused2; /* was "generation" until 2.4.0 */
1616 __u32 mbo_eadatasize;
1618 __u32 mbo_max_mdsize;
1619 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1620 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1621 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1622 __u32 mbo_padding_5; /* also fix lustre_swab_mdt_body */
1623 __u64 mbo_padding_6;
1624 __u64 mbo_padding_7;
1625 __u64 mbo_padding_8;
1626 __u64 mbo_padding_9;
1627 __u64 mbo_padding_10;
1630 struct mdt_ioepoch {
1631 struct lustre_handle mio_handle;
1632 __u64 mio_unused1; /* was ioepoch */
1633 __u32 mio_unused2; /* was flags */
1637 /* permissions for md_perm.mp_perm */
1639 CFS_SETUID_PERM = 0x01,
1640 CFS_SETGID_PERM = 0x02,
1641 CFS_SETGRP_PERM = 0x04,
1644 struct mdt_rec_setattr {
1654 __u32 sa_padding_1_h;
1655 struct lu_fid sa_fid;
1664 __u32 sa_attr_flags;
1666 __u32 sa_bias; /* some operation flags */
1673 * Attribute flags used in mdt_rec_setattr::sa_valid.
1674 * The kernel's #defines for ATTR_* should not be used over the network
1675 * since the client and MDS may run different kernels (see bug 13828)
1676 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1678 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1679 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1680 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1681 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1682 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1683 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1684 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1685 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1686 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1687 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1688 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1689 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1690 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1691 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1692 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1693 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1696 #define FMODE_READ 00000001
1697 #define FMODE_WRITE 00000002
1700 #define MDS_FMODE_CLOSED 00000000
1701 #define MDS_FMODE_EXEC 00000004
1702 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1703 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1704 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1706 #define MDS_OPEN_CREATED 00000010
1707 #define MDS_OPEN_CROSS 00000020
1709 #define MDS_OPEN_CREAT 00000100
1710 #define MDS_OPEN_EXCL 00000200
1711 #define MDS_OPEN_TRUNC 00001000
1712 #define MDS_OPEN_APPEND 00002000
1713 #define MDS_OPEN_SYNC 00010000
1714 #define MDS_OPEN_DIRECTORY 00200000
1716 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1717 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1718 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1719 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1720 * We do not support JOIN FILE
1721 * anymore, reserve this flags
1722 * just for preventing such bit
1725 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1726 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1727 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1728 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1729 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1731 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1733 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1734 * delegation, succeed if it's not
1735 * being opened with conflict mode.
1737 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1739 /* lustre internal open flags, which should not be set from user space */
1740 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1741 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1742 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1746 MDS_CHECK_SPLIT = 1 << 0,
1747 MDS_CROSS_REF = 1 << 1,
1748 MDS_VTX_BYPASS = 1 << 2,
1749 MDS_PERM_BYPASS = 1 << 3,
1750 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1751 MDS_QUOTA_IGNORE = 1 << 5,
1752 /* Was MDS_CLOSE_CLEANUP (1 << 6), No more used */
1753 MDS_KEEP_ORPHAN = 1 << 7,
1754 MDS_RECOV_OPEN = 1 << 8,
1755 MDS_DATA_MODIFIED = 1 << 9,
1756 MDS_CREATE_VOLATILE = 1 << 10,
1757 MDS_OWNEROVERRIDE = 1 << 11,
1758 MDS_HSM_RELEASE = 1 << 12,
1759 MDS_RENAME_MIGRATE = 1 << 13,
1760 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1763 /* instance of mdt_reint_rec */
1764 struct mdt_rec_create {
1772 __u32 cr_suppgid1_h;
1774 __u32 cr_suppgid2_h;
1775 struct lu_fid cr_fid1;
1776 struct lu_fid cr_fid2;
1777 struct lustre_handle cr_old_handle; /* handle in case of open replay */
1781 __u64 cr_padding_1; /* rr_blocks */
1784 /* use of helpers set/get_mrc_cr_flags() is needed to access
1785 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
1786 * extend cr_flags size without breaking 1.8 compat */
1787 __u32 cr_flags_l; /* for use with open, low 32 bits */
1788 __u32 cr_flags_h; /* for use with open, high 32 bits */
1789 __u32 cr_umask; /* umask for create */
1790 __u32 cr_padding_4; /* rr_padding_4 */
1793 /* instance of mdt_reint_rec */
1794 struct mdt_rec_link {
1802 __u32 lk_suppgid1_h;
1804 __u32 lk_suppgid2_h;
1805 struct lu_fid lk_fid1;
1806 struct lu_fid lk_fid2;
1808 __u64 lk_padding_1; /* rr_atime */
1809 __u64 lk_padding_2; /* rr_ctime */
1810 __u64 lk_padding_3; /* rr_size */
1811 __u64 lk_padding_4; /* rr_blocks */
1813 __u32 lk_padding_5; /* rr_mode */
1814 __u32 lk_padding_6; /* rr_flags */
1815 __u32 lk_padding_7; /* rr_padding_2 */
1816 __u32 lk_padding_8; /* rr_padding_3 */
1817 __u32 lk_padding_9; /* rr_padding_4 */
1820 /* instance of mdt_reint_rec */
1821 struct mdt_rec_unlink {
1829 __u32 ul_suppgid1_h;
1831 __u32 ul_suppgid2_h;
1832 struct lu_fid ul_fid1;
1833 struct lu_fid ul_fid2;
1835 __u64 ul_padding_2; /* rr_atime */
1836 __u64 ul_padding_3; /* rr_ctime */
1837 __u64 ul_padding_4; /* rr_size */
1838 __u64 ul_padding_5; /* rr_blocks */
1841 __u32 ul_padding_6; /* rr_flags */
1842 __u32 ul_padding_7; /* rr_padding_2 */
1843 __u32 ul_padding_8; /* rr_padding_3 */
1844 __u32 ul_padding_9; /* rr_padding_4 */
1847 /* instance of mdt_reint_rec */
1848 struct mdt_rec_rename {
1856 __u32 rn_suppgid1_h;
1858 __u32 rn_suppgid2_h;
1859 struct lu_fid rn_fid1;
1860 struct lu_fid rn_fid2;
1862 __u64 rn_padding_1; /* rr_atime */
1863 __u64 rn_padding_2; /* rr_ctime */
1864 __u64 rn_padding_3; /* rr_size */
1865 __u64 rn_padding_4; /* rr_blocks */
1866 __u32 rn_bias; /* some operation flags */
1867 __u32 rn_mode; /* cross-ref rename has mode */
1868 __u32 rn_padding_5; /* rr_flags */
1869 __u32 rn_padding_6; /* rr_padding_2 */
1870 __u32 rn_padding_7; /* rr_padding_3 */
1871 __u32 rn_padding_8; /* rr_padding_4 */
1874 /* instance of mdt_reint_rec */
1875 struct mdt_rec_setxattr {
1883 __u32 sx_suppgid1_h;
1885 __u32 sx_suppgid2_h;
1886 struct lu_fid sx_fid;
1887 __u64 sx_padding_1; /* These three are rr_fid2 */
1892 __u64 sx_padding_5; /* rr_ctime */
1893 __u64 sx_padding_6; /* rr_size */
1894 __u64 sx_padding_7; /* rr_blocks */
1897 __u32 sx_padding_8; /* rr_flags */
1898 __u32 sx_padding_9; /* rr_padding_2 */
1899 __u32 sx_padding_10; /* rr_padding_3 */
1900 __u32 sx_padding_11; /* rr_padding_4 */
1904 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
1905 * Do NOT change the size of various members, otherwise the value
1906 * will be broken in lustre_swab_mdt_rec_reint().
1908 * If you add new members in other mdt_reint_xxx structres and need to use the
1909 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
1911 struct mdt_rec_reint {
1919 __u32 rr_suppgid1_h;
1921 __u32 rr_suppgid2_h;
1922 struct lu_fid rr_fid1;
1923 struct lu_fid rr_fid2;
1934 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1937 /* lmv structures */
1939 __u32 ld_tgt_count; /* how many MDS's */
1940 __u32 ld_active_tgt_count; /* how many active */
1941 __u32 ld_default_stripe_count; /* how many objects are used */
1942 __u32 ld_pattern; /* default hash pattern */
1943 __u64 ld_default_hash_size;
1944 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
1945 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
1946 __u32 ld_qos_maxage; /* in second */
1947 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
1948 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
1949 struct obd_uuid ld_uuid;
1952 /* LMV layout EA, and it will be stored both in master and slave object */
1953 struct lmv_mds_md_v1 {
1955 __u32 lmv_stripe_count;
1956 __u32 lmv_master_mdt_index; /* On master object, it is master
1957 * MDT index, on slave object, it
1958 * is stripe index of the slave obj */
1959 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
1960 * which hash function to be used,
1961 * Note: only lower 16 bits is being
1962 * used for now. Higher 16 bits will
1963 * be used to mark the object status,
1964 * for example migrating or dead. */
1965 __u32 lmv_layout_version; /* Used for directory restriping */
1969 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
1970 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
1973 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
1974 #define LMV_MAGIC LMV_MAGIC_V1
1976 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
1977 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
1979 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
1980 * and the higher part will be the flag to indicate the status of object,
1981 * for example the object is being migrated. And the hash function
1982 * might be interpreted differently with different flags. */
1983 #define LMV_HASH_TYPE_MASK 0x0000ffff
1985 #define LMV_HASH_FLAG_MIGRATION 0x80000000
1987 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 10, 53, 0)
1988 /* Since lustre 2.8, this flag will not be needed, instead this DEAD
1989 * and orphan flags will be stored in LMA (see LMAI_ORPHAN)
1990 * Keep this flag just for LFSCK, because it still might meet such
1991 * flag when it checks the old FS */
1992 #define LMV_HASH_FLAG_DEAD 0x40000000
1994 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
1996 /* The striped directory has ever lost its master LMV EA, then LFSCK
1997 * re-generated it. This flag is used to indicate such case. It is an
1999 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
2002 * The FNV-1a hash algorithm is as follows:
2003 * hash = FNV_offset_basis
2004 * for each octet_of_data to be hashed
2005 * hash = hash XOR octet_of_data
2006 * hash = hash × FNV_prime
2008 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2010 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2011 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2013 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2014 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2015 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, size_t size)
2017 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2018 const unsigned char *p = buf;
2021 for (i = 0; i < size; i++) {
2023 hash *= LUSTRE_FNV_1A_64_PRIME;
2031 struct lmv_mds_md_v1 lmv_md_v1;
2032 struct lmv_user_md lmv_user_md;
2035 static inline int lmv_mds_md_size(int stripe_count, unsigned int lmm_magic)
2037 switch (lmm_magic) {
2039 struct lmv_mds_md_v1 *lmm1;
2041 return sizeof(*lmm1) + stripe_count *
2042 sizeof(lmm1->lmv_stripe_fids[0]);
2049 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2051 switch (__le32_to_cpu(lmm->lmv_magic)) {
2053 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2054 case LMV_USER_MAGIC:
2055 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2061 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2063 switch (__le32_to_cpu(lmm->lmv_magic)) {
2065 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2066 case LMV_USER_MAGIC:
2067 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2077 FLD_FIRST_OPC = FLD_QUERY
2083 SEQ_FIRST_OPC = SEQ_QUERY
2087 SEQ_ALLOC_SUPER = 0,
2099 LFSCK_NOTIFY = 1101,
2102 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2106 * LOV data structures
2109 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2110 /* The size of the buffer the lov/mdc reserves for the
2111 * array of UUIDs returned by the MDS. With the current
2112 * protocol, this will limit the max number of OSTs per LOV */
2114 #define LOV_DESC_MAGIC 0xB0CCDE5C
2115 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2116 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2118 /* LOV settings descriptor (should only contain static info) */
2120 __u32 ld_tgt_count; /* how many OBD's */
2121 __u32 ld_active_tgt_count; /* how many active */
2122 __u32 ld_default_stripe_count; /* how many objects are used */
2123 __u32 ld_pattern; /* default PATTERN_RAID0 */
2124 __u64 ld_default_stripe_size; /* in bytes */
2125 __u64 ld_default_stripe_offset; /* in bytes */
2126 __u32 ld_padding_0; /* unused */
2127 __u32 ld_qos_maxage; /* in second */
2128 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2129 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2130 struct obd_uuid ld_uuid;
2133 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2138 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2143 LDLM_BL_CALLBACK = 104,
2144 LDLM_CP_CALLBACK = 105,
2145 LDLM_GL_CALLBACK = 106,
2146 LDLM_SET_INFO = 107,
2149 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2151 #define RES_NAME_SIZE 4
2152 struct ldlm_res_id {
2153 __u64 name[RES_NAME_SIZE];
2156 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2157 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2158 (unsigned long long)(res)->lr_name.name[1], \
2159 (unsigned long long)(res)->lr_name.name[2], \
2160 (unsigned long long)(res)->lr_name.name[3]
2163 typedef enum ldlm_mode {
2176 #define LCK_MODE_NUM 8
2178 typedef enum ldlm_type {
2186 #define LDLM_MIN_TYPE LDLM_PLAIN
2188 struct ldlm_extent {
2194 struct ldlm_inodebits {
2198 struct ldlm_flock_wire {
2206 /* it's important that the fields of the ldlm_extent structure match
2207 * the first fields of the ldlm_flock structure because there is only
2208 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2209 * this ever changes we will need to swab the union differently based
2210 * on the resource type. */
2212 typedef union ldlm_wire_policy_data {
2213 struct ldlm_extent l_extent;
2214 struct ldlm_flock_wire l_flock;
2215 struct ldlm_inodebits l_inodebits;
2216 } ldlm_wire_policy_data_t;
2218 union ldlm_gl_desc {
2219 struct ldlm_gl_lquota_desc lquota_desc;
2222 enum ldlm_intent_flags {
2223 IT_OPEN = 0x00000001,
2224 IT_CREAT = 0x00000002,
2225 IT_OPEN_CREAT = 0x00000003,
2226 IT_READDIR = 0x00000004,
2227 IT_GETATTR = 0x00000008,
2228 IT_LOOKUP = 0x00000010,
2229 IT_UNLINK = 0x00000020,
2230 IT_TRUNC = 0x00000040,
2231 IT_GETXATTR = 0x00000080,
2232 IT_EXEC = 0x00000100,
2233 IT_PIN = 0x00000200,
2234 IT_LAYOUT = 0x00000400,
2235 IT_QUOTA_DQACQ = 0x00000800,
2236 IT_QUOTA_CONN = 0x00001000,
2237 IT_SETXATTR = 0x00002000,
2240 struct ldlm_intent {
2244 struct ldlm_resource_desc {
2245 enum ldlm_type lr_type;
2246 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2247 struct ldlm_res_id lr_name;
2250 struct ldlm_lock_desc {
2251 struct ldlm_resource_desc l_resource;
2252 enum ldlm_mode l_req_mode;
2253 enum ldlm_mode l_granted_mode;
2254 union ldlm_wire_policy_data l_policy_data;
2257 #define LDLM_LOCKREQ_HANDLES 2
2258 #define LDLM_ENQUEUE_CANCEL_OFF 1
2260 struct ldlm_request {
2263 struct ldlm_lock_desc lock_desc;
2264 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2269 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2270 struct ldlm_lock_desc lock_desc;
2271 struct lustre_handle lock_handle;
2272 __u64 lock_policy_res1;
2273 __u64 lock_policy_res2;
2276 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2277 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2280 * Opcodes for mountconf (mgs and mgc)
2285 MGS_EXCEPTION, /* node died, etc. */
2286 MGS_TARGET_REG, /* whenever target starts up */
2292 #define MGS_FIRST_OPC MGS_CONNECT
2294 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 13, 53, 0)
2295 #define MGS_PARAM_MAXLEN 1024
2296 #define KEY_SET_INFO "set_info"
2298 struct mgs_send_param {
2299 char mgs_param[MGS_PARAM_MAXLEN];
2303 /* We pass this info to the MGS so it can write config logs */
2304 #define MTI_NAME_MAXLEN 64
2305 #define MTI_PARAM_MAXLEN 4096
2306 #define MTI_NIDS_MAX 32
2307 struct mgs_target_info {
2308 __u32 mti_lustre_ver;
2309 __u32 mti_stripe_index;
2310 __u32 mti_config_ver;
2312 __u32 mti_nid_count;
2313 __u32 mti_instance; /* Running instance of target */
2314 char mti_fsname[MTI_NAME_MAXLEN];
2315 char mti_svname[MTI_NAME_MAXLEN];
2316 char mti_uuid[sizeof(struct obd_uuid)];
2317 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
2318 char mti_params[MTI_PARAM_MAXLEN];
2321 struct mgs_nidtbl_entry {
2322 __u64 mne_version; /* table version of this entry */
2323 __u32 mne_instance; /* target instance # */
2324 __u32 mne_index; /* target index */
2325 __u32 mne_length; /* length of this entry - by bytes */
2326 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2327 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2328 __u8 mne_nid_size; /* size of each NID, by bytes */
2329 __u8 mne_nid_count; /* # of NIDs in buffer */
2331 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2335 struct mgs_config_body {
2336 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2337 __u64 mcb_offset; /* next index of config log to request */
2338 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2339 __u8 mcb_nm_cur_pass;
2340 __u8 mcb_bits; /* bits unit size of config log */
2341 __u32 mcb_units; /* # of units for bulk transfer */
2344 struct mgs_config_res {
2345 __u64 mcr_offset; /* index of last config log */
2347 __u64 mcr_size; /* size of the log */
2348 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2352 /* Config marker flags (in config log) */
2353 #define CM_START 0x01
2355 #define CM_SKIP 0x04
2356 #define CM_UPGRADE146 0x08
2357 #define CM_EXCLUDE 0x10
2358 #define CM_START_SKIP (CM_START | CM_SKIP)
2361 __u32 cm_step; /* aka config version */
2363 __u32 cm_vers; /* lustre release version number */
2364 __u32 cm_padding; /* 64 bit align */
2365 __s64 cm_createtime; /*when this record was first created */
2366 __s64 cm_canceltime; /*when this record is no longer valid*/
2367 char cm_tgtname[MTI_NAME_MAXLEN];
2368 char cm_comment[MTI_NAME_MAXLEN];
2372 * Opcodes for multiple servers.
2378 OBD_QC_CALLBACK, /* not used since 2.4 */
2382 #define OBD_FIRST_OPC OBD_PING
2385 * llog contexts indices.
2387 * There is compatibility problem with indexes below, they are not
2388 * continuous and must keep their numbers for compatibility needs.
2389 * See LU-5218 for details.
2392 LLOG_CONFIG_ORIG_CTXT = 0,
2393 LLOG_CONFIG_REPL_CTXT = 1,
2394 LLOG_MDS_OST_ORIG_CTXT = 2,
2395 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2396 LLOG_SIZE_ORIG_CTXT = 4,
2397 LLOG_SIZE_REPL_CTXT = 5,
2398 LLOG_TEST_ORIG_CTXT = 8,
2399 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2400 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2401 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2402 /* for multiple changelog consumers */
2403 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2404 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2405 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log */
2406 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log */
2410 /** Identifier for a single log object */
2412 struct ost_id lgl_oi;
2414 } __attribute__((packed));
2416 /** Records written to the CATALOGS list */
2417 #define CATLIST "CATALOGS"
2419 struct llog_logid lci_logid;
2423 } __attribute__((packed));
2425 /* Log data record types - there is no specific reason that these need to
2426 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2428 #define LLOG_OP_MAGIC 0x10600000
2429 #define LLOG_OP_MASK 0xfff00000
2432 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2433 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2434 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2435 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2436 REINT_UNLINK, /* obsolete after 2.5.0 */
2437 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2439 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2440 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2442 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2443 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2444 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2445 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2446 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2447 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2448 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2449 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000,
2450 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2451 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2454 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2455 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2457 /** Log record header - stored in little endian order.
2458 * Each record must start with this struct, end with a llog_rec_tail,
2459 * and be a multiple of 256 bits in size.
2461 struct llog_rec_hdr {
2468 struct llog_rec_tail {
2473 /* Where data follow just after header */
2474 #define REC_DATA(ptr) \
2475 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2477 #define REC_DATA_LEN(rec) \
2478 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2479 sizeof(struct llog_rec_tail))
2481 struct llog_logid_rec {
2482 struct llog_rec_hdr lid_hdr;
2483 struct llog_logid lid_id;
2487 struct llog_rec_tail lid_tail;
2488 } __attribute__((packed));
2490 struct llog_unlink_rec {
2491 struct llog_rec_hdr lur_hdr;
2495 struct llog_rec_tail lur_tail;
2496 } __attribute__((packed));
2498 struct llog_unlink64_rec {
2499 struct llog_rec_hdr lur_hdr;
2500 struct lu_fid lur_fid;
2501 __u32 lur_count; /* to destroy the lost precreated */
2505 struct llog_rec_tail lur_tail;
2506 } __attribute__((packed));
2508 struct llog_setattr64_rec {
2509 struct llog_rec_hdr lsr_hdr;
2510 struct ost_id lsr_oi;
2516 struct llog_rec_tail lsr_tail;
2517 } __attribute__((packed));
2519 struct llog_size_change_rec {
2520 struct llog_rec_hdr lsc_hdr;
2521 struct ll_fid lsc_fid;
2526 struct llog_rec_tail lsc_tail;
2527 } __attribute__((packed));
2529 #define CHANGELOG_MAGIC 0xca103000
2531 /** \a changelog_rec_type's that can't be masked */
2532 #define CHANGELOG_MINMASK (1 << CL_MARK)
2533 /** bits covering all \a changelog_rec_type's */
2534 #define CHANGELOG_ALLMASK 0XFFFFFFFF
2535 /** default \a changelog_rec_type mask. Allow all of them, except
2536 * CL_ATIME since it can really be time consuming, and not necessary
2537 * under normal use. */
2538 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & ~(1 << CL_ATIME))
2540 /* changelog llog name, needed by client replicators */
2541 #define CHANGELOG_CATALOG "changelog_catalog"
2543 struct changelog_setinfo {
2546 } __attribute__((packed));
2548 /** changelog record */
2549 struct llog_changelog_rec {
2550 struct llog_rec_hdr cr_hdr;
2551 struct changelog_rec cr; /**< Variable length field */
2552 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2553 } __attribute__((packed));
2555 #define CHANGELOG_USER_PREFIX "cl"
2557 struct llog_changelog_user_rec {
2558 struct llog_rec_hdr cur_hdr;
2562 struct llog_rec_tail cur_tail;
2563 } __attribute__((packed));
2565 enum agent_req_status {
2573 static inline const char *agent_req_status2name(enum agent_req_status ars)
2591 struct llog_agent_req_rec {
2592 struct llog_rec_hdr arr_hdr; /**< record header */
2593 __u32 arr_status; /**< status of the request */
2595 * agent_req_status */
2596 __u32 arr_archive_id; /**< backend archive number */
2597 __u64 arr_flags; /**< req flags */
2598 __u64 arr_compound_id; /**< compound cookie */
2599 __u64 arr_req_create; /**< req. creation time */
2600 __u64 arr_req_change; /**< req. status change time */
2601 struct hsm_action_item arr_hai; /**< req. to the agent */
2602 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2603 } __attribute__((packed));
2605 /* Old llog gen for compatibility */
2609 } __attribute__((packed));
2611 struct llog_gen_rec {
2612 struct llog_rec_hdr lgr_hdr;
2613 struct llog_gen lgr_gen;
2617 struct llog_rec_tail lgr_tail;
2620 /* flags for the logs */
2622 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2623 LLOG_F_IS_CAT = 0x2,
2624 LLOG_F_IS_PLAIN = 0x4,
2625 LLOG_F_EXT_JOBID = 0x8,
2626 LLOG_F_IS_FIXSIZE = 0x10,
2628 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
2629 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
2630 * because the catlog record is usually fixed size, but its plain
2631 * log record can be variable */
2632 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID,
2635 /* On-disk header structure of each log object, stored in little endian order */
2636 #define LLOG_MIN_CHUNK_SIZE 8192
2637 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
2638 * - sizeof(llh_bitmap) */
2639 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
2640 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2642 struct llog_log_hdr {
2643 struct llog_rec_hdr llh_hdr;
2644 __s64 llh_timestamp;
2646 __u32 llh_bitmap_offset;
2649 /* for a catalog the first/oldest and still in-use plain slot is just
2650 * next to it. It will serve as the upper limit after Catalog has
2653 struct obd_uuid llh_tgtuuid;
2654 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
2655 /* These fields must always be at the end of the llog_log_hdr.
2656 * Note: llh_bitmap size is variable because llog chunk size could be
2657 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
2658 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
2659 * llh_tail should only be refered by LLOG_HDR_TAIL().
2660 * But this structure is also used by client/server llog interface
2661 * (see llog_client.c), it will be kept in its original way to avoid
2662 * compatiblity issue. */
2663 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
2664 struct llog_rec_tail llh_tail;
2665 } __attribute__((packed));
2666 #undef LLOG_HEADER_SIZE
2667 #undef LLOG_BITMAP_BYTES
2669 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2670 llh->llh_bitmap_offset - \
2671 sizeof(llh->llh_tail)) * 8)
2672 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
2673 (llh)->llh_bitmap_offset)
2674 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
2675 llh->llh_hdr.lrh_len - \
2676 sizeof(llh->llh_tail)))
2678 /** log cookies are used to reference a specific log file and a record therein */
2679 struct llog_cookie {
2680 struct llog_logid lgc_lgl;
2684 } __attribute__((packed));
2686 /** llog protocol */
2687 enum llogd_rpc_ops {
2688 LLOG_ORIGIN_HANDLE_CREATE = 501,
2689 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2690 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2691 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2692 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2693 LLOG_ORIGIN_CONNECT = 506,
2694 LLOG_CATINFO = 507, /* deprecated */
2695 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2696 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2698 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2702 struct llog_logid lgd_logid;
2704 __u32 lgd_llh_flags;
2706 __u32 lgd_saved_index;
2708 __u64 lgd_cur_offset;
2709 } __attribute__((packed));
2711 struct llogd_conn_body {
2712 struct llog_gen lgdc_gen;
2713 struct llog_logid lgdc_logid;
2714 __u32 lgdc_ctxt_idx;
2715 } __attribute__((packed));
2717 /* Note: 64-bit types are 64-bit aligned in structure */
2719 __u64 o_valid; /* hot fields in this obdo */
2722 __u64 o_size; /* o_size-o_blocks == ost_lvb */
2726 __u64 o_blocks; /* brw: cli sent cached bytes */
2729 /* 32-bit fields start here: keep an even number of them via padding */
2730 __u32 o_blksize; /* optimal IO blocksize */
2731 __u32 o_mode; /* brw: cli sent cache remain */
2735 __u32 o_nlink; /* brw: checksum */
2737 __u32 o_misc; /* brw: o_dropped */
2739 __u64 o_ioepoch; /* epoch in ost writes */
2740 __u32 o_stripe_idx; /* holds stripe idx */
2742 struct lustre_handle o_handle; /* brw: lock handle to prolong
2744 struct llog_cookie o_lcookie; /* destroy: unlink cookie from
2745 * MDS, obsolete in 2.8, reused
2750 __u64 o_data_version; /* getattr: sum of iversion for
2752 * brw: grant space consumed on
2753 * the client for the write */
2759 #define o_dirty o_blocks
2760 #define o_undirty o_mode
2761 #define o_dropped o_misc
2762 #define o_cksum o_nlink
2763 #define o_grant_used o_data_version
2765 struct lfsck_request {
2778 __u16 lr_async_windows;
2780 struct lu_fid lr_fid;
2781 struct lu_fid lr_fid2;
2782 struct lu_fid lr_fid3;
2787 struct lfsck_reply {
2794 LE_LASTID_REBUILDING = 1,
2795 LE_LASTID_REBUILT = 2,
2801 LE_FID_ACCESSED = 8,
2803 LE_CONDITIONAL_DESTROY = 10,
2804 LE_PAIRS_VERIFY = 11,
2805 LE_SET_LMV_MASTER = 15,
2806 LE_SET_LMV_SLAVE = 16,
2809 enum lfsck_event_flags {
2810 LEF_TO_OST = 0x00000001,
2811 LEF_FROM_OST = 0x00000002,
2812 LEF_SET_LMV_HASH = 0x00000004,
2813 LEF_SET_LMV_ALL = 0x00000008,
2814 LEF_RECHECK_NAME_HASH = 0x00000010,
2815 LEF_QUERY_ALL = 0x00000020,
2818 /* request structure for OST's */
2823 /* Key for FIEMAP to be used in get_info calls */
2824 struct ll_fiemap_info_key {
2826 struct obdo lfik_oa;
2827 struct fiemap lfik_fiemap;
2830 #define IDX_INFO_MAGIC 0x3D37CC37
2832 /* Index file transfer through the network. The server serializes the index into
2833 * a byte stream which is sent to the client via a bulk transfer */
2837 /* reply: see idx_info_flags below */
2840 /* request & reply: number of lu_idxpage (to be) transferred */
2844 /* request: requested attributes passed down to the iterator API */
2847 /* request & reply: index file identifier (FID) */
2848 struct lu_fid ii_fid;
2850 /* reply: version of the index file before starting to walk the index.
2851 * Please note that the version can be modified at any time during the
2855 /* request: hash to start with:
2856 * reply: hash of the first entry of the first lu_idxpage and hash
2857 * of the entry to read next if any */
2858 __u64 ii_hash_start;
2861 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
2865 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
2874 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
2876 /* List of flags used in idx_info::ii_flags */
2877 enum idx_info_flags {
2878 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
2879 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
2880 II_FL_VARREC = 1 << 2, /* records can be of variable size */
2881 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
2882 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
2885 #define LIP_MAGIC 0x8A6D6B6C
2887 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
2889 /* 16-byte header */
2892 __u16 lip_nr; /* number of entries in the container */
2893 __u64 lip_pad0; /* additional padding for future use */
2895 /* key/record pairs are stored in the remaining 4080 bytes.
2896 * depending upon the flags in idx_info::ii_flags, each key/record
2897 * pair might be preceded by:
2899 * - the key size (II_FL_VARKEY is set)
2900 * - the record size (II_FL_VARREC is set)
2902 * For the time being, we only support fixed-size key & record. */
2903 char lip_entries[0];
2906 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
2908 /* Gather all possible type associated with a 4KB container */
2910 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
2911 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
2912 char lp_array[LU_PAGE_SIZE];
2915 /* security opcodes */
2918 SEC_CTX_INIT_CONT = 802,
2921 SEC_FIRST_OPC = SEC_CTX_INIT
2925 * capa related definitions
2927 #define CAPA_HMAC_MAX_LEN 64
2928 #define CAPA_HMAC_KEY_MAX_LEN 56
2930 /* NB take care when changing the sequence of elements this struct,
2931 * because the offset info is used in find_capa() */
2932 struct lustre_capa {
2933 struct lu_fid lc_fid; /** fid */
2934 __u64 lc_opc; /** operations allowed */
2935 __u64 lc_uid; /** file owner */
2936 __u64 lc_gid; /** file group */
2937 __u32 lc_flags; /** HMAC algorithm & flags */
2938 __u32 lc_keyid; /** key# used for the capability */
2939 __u32 lc_timeout; /** capa timeout value (sec) */
2940 __u32 lc_expiry; /** expiry time (sec) */
2941 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
2942 } __attribute__((packed));
2944 /** lustre_capa::lc_opc */
2946 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
2947 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
2948 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
2949 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
2950 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
2951 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
2952 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
2953 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
2954 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
2955 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
2956 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
2959 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2960 #define CAPA_OPC_MDS_ONLY \
2961 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2962 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2963 #define CAPA_OPC_OSS_ONLY \
2964 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
2965 CAPA_OPC_OSS_DESTROY)
2966 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2967 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2969 /* lustre_capa::lc_hmac_alg */
2971 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
2975 #define CAPA_FL_MASK 0x00ffffff
2976 #define CAPA_HMAC_ALG_MASK 0xff000000
2978 struct lustre_capa_key {
2979 __u64 lk_seq; /**< mds# */
2980 __u32 lk_keyid; /**< key# */
2982 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
2983 } __attribute__((packed));
2985 /** The link ea holds 1 \a link_ea_entry for each hardlink */
2986 #define LINK_EA_MAGIC 0x11EAF1DFUL
2987 struct link_ea_header {
2990 __u64 leh_len; /* total size */
2991 __u32 leh_overflow_time;
2995 /** Hardlink data is name and parent fid.
2996 * Stored in this crazy struct for maximum packing and endian-neutrality
2998 struct link_ea_entry {
2999 /** __u16 stored big-endian, unaligned */
3000 unsigned char lee_reclen[2];
3001 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3003 }__attribute__((packed));
3005 /** fid2path request/reply structure */
3006 struct getinfo_fid2path {
3007 struct lu_fid gf_fid;
3013 struct lu_fid gf_root_fid[0];
3015 } __attribute__((packed));
3017 /** path2parent request/reply structures */
3019 struct lu_fid gp_fid; /**< parent FID */
3020 __u32 gp_linkno; /**< hardlink number */
3021 __u32 gp_name_size; /**< size of the name field */
3022 char gp_name[0]; /**< zero-terminated link name */
3023 } __attribute__((packed));
3026 LAYOUT_INTENT_ACCESS = 0,
3027 LAYOUT_INTENT_READ = 1,
3028 LAYOUT_INTENT_WRITE = 2,
3029 LAYOUT_INTENT_GLIMPSE = 3,
3030 LAYOUT_INTENT_TRUNC = 4,
3031 LAYOUT_INTENT_RELEASE = 5,
3032 LAYOUT_INTENT_RESTORE = 6
3035 /* enqueue layout lock with intent */
3036 struct layout_intent {
3037 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3044 * On the wire version of hsm_progress structure.
3046 * Contains the userspace hsm_progress and some internal fields.
3048 struct hsm_progress_kernel {
3049 /* Field taken from struct hsm_progress */
3052 struct hsm_extent hpk_extent;
3054 __u16 hpk_errval; /* positive val */
3056 /* Additional fields */
3057 __u64 hpk_data_version;
3059 } __attribute__((packed));
3062 * OUT_UPDATE RPC Format
3064 * During the cross-ref operation, the Master MDT, which the client send the
3065 * request to, will disassembly the operation into object updates, then OSP
3066 * will send these updates to the remote MDT to be executed.
3068 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3069 * operation and does a type of modification to an object.
3077 * update (ub_count-th)
3079 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3084 * rc [+ buffers] (1st)
3085 * rc [+ buffers] (2st)
3087 * rc [+ buffers] (nr_count-th)
3089 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3090 * be equal to ub_count.
3094 * Type of each update, if adding/deleting update, please also update
3095 * update_opcode in lustre/target/out_lib.c.
3107 OUT_INDEX_LOOKUP = 9,
3108 OUT_INDEX_INSERT = 10,
3109 OUT_INDEX_DELETE = 11,
3119 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3120 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3121 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3122 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3125 struct object_update_param {
3126 __u16 oup_len; /* length of this parameter */
3133 struct object_update {
3134 __u16 ou_type; /* enum update_type */
3135 __u16 ou_params_count; /* update parameters count */
3136 __u32 ou_result_size; /* how many bytes can return */
3137 __u32 ou_flags; /* enum update_flag */
3138 __u32 ou_padding1; /* padding 1 */
3139 __u64 ou_batchid; /* op transno on master */
3140 struct lu_fid ou_fid; /* object to be updated */
3141 struct object_update_param ou_params[0]; /* update params */
3144 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3145 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3146 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3147 /* Hold object_updates sending to the remote OUT in single RPC */
3148 struct object_update_request {
3150 __u16 ourq_count; /* number of ourq_updates[] */
3152 struct object_update ourq_updates[0];
3155 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3156 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3157 /* Header for updates request between MDTs */
3158 struct out_update_header {
3161 __u32 ouh_inline_length;
3162 __u32 ouh_reply_size;
3163 __u32 ouh_inline_data[0];
3166 struct out_update_buffer {
3171 /* the result of object update */
3172 struct object_update_result {
3179 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3180 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3181 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3182 /* Hold object_update_results being replied from the remote OUT. */
3183 struct object_update_reply {
3190 /* read update result */
3191 struct out_read_reply {
3198 /** layout swap request structure
3199 * fid1 and fid2 are in mdt_body
3201 struct mdc_swap_layouts {
3206 struct lustre_handle cd_handle;
3207 struct lu_fid cd_fid;
3208 __u64 cd_data_version;
3209 __u64 cd_reserved[8];
3212 /* Update llog format */
3214 struct lu_fid uop_fid;
3216 __u16 uop_param_count;
3217 __u16 uop_params_off[0];
3221 struct update_op uops_op[0];
3224 struct update_params {
3225 struct object_update_param up_params[0];
3228 enum update_records_flag {
3229 UPDATE_RECORD_CONTINUE = 1 >> 0,
3232 * This is the update record format used to store the updates in
3233 * disk. All updates of the operation will be stored in ur_ops.
3234 * All of parameters for updates of the operation will be stored
3236 * To save the space of the record, parameters in ur_ops will only
3237 * remember their offset in ur_params, so to avoid storing duplicate
3238 * parameters in ur_params, which can help us save a lot space for
3239 * operation like creating striped directory.
3241 struct update_records {
3242 __u64 ur_master_transno;
3245 /* If the operation includes multiple updates, then ur_index
3246 * means the index of the update inside the whole updates. */
3248 __u32 ur_update_count;
3249 __u32 ur_param_count;
3250 struct update_ops ur_ops;
3251 /* Note ur_ops has a variable size, so comment out
3252 * the following ur_params, in case some use it directly
3253 * update_records->ur_params
3255 * struct update_params ur_params;
3259 struct llog_update_record {
3260 struct llog_rec_hdr lur_hdr;
3261 struct update_records lur_update_rec;
3262 /* Note ur_update_rec has a variable size, so comment out
3263 * the following ur_tail, in case someone use it directly
3265 * struct llog_rec_tail lur_tail;
3269 /* nodemap records, uses 32 byte record length */
3270 #define LUSTRE_NODEMAP_NAME_LENGTH 16
3271 struct nodemap_cluster_rec {
3272 char ncr_name[LUSTRE_NODEMAP_NAME_LENGTH + 1];
3276 __u32 ncr_squash_uid;
3277 __u32 ncr_squash_gid;
3280 /* lnet_nid_t is 8 bytes */
3281 struct nodemap_range_rec {
3282 lnet_nid_t nrr_start_nid;
3283 lnet_nid_t nrr_end_nid;
3288 struct nodemap_id_rec {
3296 struct nodemap_global_rec {
3307 struct nodemap_cluster_rec ncr;
3308 struct nodemap_range_rec nrr;
3309 struct nodemap_id_rec nir;
3310 struct nodemap_global_rec ngr;
3313 /* This is the lu_ladvise struct which goes out on the wire.
3314 * Corresponds to the userspace arg llapi_lu_ladvise.
3315 * value[1-4] are unspecified fields, used differently by different advices */
3317 __u16 lla_advice; /* advice type */
3318 __u16 lla_value1; /* values for different advice types */
3320 __u64 lla_start; /* first byte of extent for advice */
3321 __u64 lla_end; /* last byte of extent for advice */
3326 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3327 * arg llapi_ladvise_hdr.
3328 * value[1-3] are unspecified fields, used differently by different advices */
3329 struct ladvise_hdr {
3330 __u32 lah_magic; /* LADVISE_MAGIC */
3331 __u32 lah_count; /* number of advices */
3332 __u64 lah_flags; /* from enum ladvise_flag */
3333 __u32 lah_value1; /* unused */
3334 __u32 lah_value2; /* unused */
3335 __u64 lah_value3; /* unused */
3336 struct lu_ladvise lah_advise[0]; /* advices in this header */