4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2015, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/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, and the
52 * prototypes of the swabber functions for each struct. Nothing that
53 * depends on external functions or definitions should be in here.
55 * Structs must be properly aligned to put 64-bit values on an 8-byte
56 * boundary. Any structs being added here must also be added to
57 * utils/wirecheck.c and "make newwiretest" run to regenerate the
58 * utils/wiretest.c sources. This allows us to verify that wire structs
59 * have the proper alignment/size on all architectures.
61 * DO NOT CHANGE any of the structs, flags, values declared here and used
62 * in released Lustre versions. Some structs may have padding fields that
63 * can be used. Some structs might allow addition at the end (verify this
64 * in the code to ensure that new/old clients that see this larger struct
65 * do not fail, otherwise you need to implement protocol compatibility).
67 * We assume all nodes are either little-endian or big-endian, and we
68 * always send messages in the sender's native format. The receiver
69 * detects the message format by checking the 'magic' field of the message
70 * (see lustre_msg_swabbed() below).
72 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
73 * implemented either here, inline (trivial implementations) or in
74 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
75 * endian, in-place in the message buffer.
77 * A swabber takes a single pointer argument. The caller must already have
78 * verified that the length of the message buffer >= sizeof (type).
80 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
81 * may be defined that swabs just the variable part, after the caller has
82 * verified that the message buffer is large enough.
87 #ifndef _LUSTRE_IDL_H_
88 #define _LUSTRE_IDL_H_
90 #include "../../../include/linux/libcfs/libcfs.h"
91 #include "../../../include/linux/lnet/types.h"
93 /* Defn's shared with user-space. */
94 #include "lustre_user.h"
95 #include "lustre_errno.h"
96 #include "../lustre_ver.h"
101 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
102 * FOO_REPLY_PORTAL is for incoming replies on the FOO
103 * FOO_BULK_PORTAL is for incoming bulk on the FOO
106 /* Lustre service names are following the format
107 * service name + MDT + seq name
109 #define LUSTRE_MDT_MAXNAMELEN 80
111 #define CONNMGR_REQUEST_PORTAL 1
112 #define CONNMGR_REPLY_PORTAL 2
113 /*#define OSC_REQUEST_PORTAL 3 */
114 #define OSC_REPLY_PORTAL 4
115 /*#define OSC_BULK_PORTAL 5 */
116 #define OST_IO_PORTAL 6
117 #define OST_CREATE_PORTAL 7
118 #define OST_BULK_PORTAL 8
119 /*#define MDC_REQUEST_PORTAL 9 */
120 #define MDC_REPLY_PORTAL 10
121 /*#define MDC_BULK_PORTAL 11 */
122 #define MDS_REQUEST_PORTAL 12
123 /*#define MDS_REPLY_PORTAL 13 */
124 #define MDS_BULK_PORTAL 14
125 #define LDLM_CB_REQUEST_PORTAL 15
126 #define LDLM_CB_REPLY_PORTAL 16
127 #define LDLM_CANCEL_REQUEST_PORTAL 17
128 #define LDLM_CANCEL_REPLY_PORTAL 18
129 /*#define PTLBD_REQUEST_PORTAL 19 */
130 /*#define PTLBD_REPLY_PORTAL 20 */
131 /*#define PTLBD_BULK_PORTAL 21 */
132 #define MDS_SETATTR_PORTAL 22
133 #define MDS_READPAGE_PORTAL 23
134 #define OUT_PORTAL 24
136 #define MGC_REPLY_PORTAL 25
137 #define MGS_REQUEST_PORTAL 26
138 #define MGS_REPLY_PORTAL 27
139 #define OST_REQUEST_PORTAL 28
140 #define FLD_REQUEST_PORTAL 29
141 #define SEQ_METADATA_PORTAL 30
142 #define SEQ_DATA_PORTAL 31
143 #define SEQ_CONTROLLER_PORTAL 32
144 #define MGS_BULK_PORTAL 33
146 /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com,
151 #define PTL_RPC_MSG_REQUEST 4711
152 #define PTL_RPC_MSG_ERR 4712
153 #define PTL_RPC_MSG_REPLY 4713
155 /* DON'T use swabbed values of MAGIC as magic! */
156 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
157 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
159 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
161 #define PTLRPC_MSG_VERSION 0x00000003
162 #define LUSTRE_VERSION_MASK 0xffff0000
163 #define LUSTRE_OBD_VERSION 0x00010000
164 #define LUSTRE_MDS_VERSION 0x00020000
165 #define LUSTRE_OST_VERSION 0x00030000
166 #define LUSTRE_DLM_VERSION 0x00040000
167 #define LUSTRE_LOG_VERSION 0x00050000
168 #define LUSTRE_MGS_VERSION 0x00060000
171 * Describes a range of sequence, lsr_start is included but lsr_end is
173 * Same structure is used in fld module where lsr_index field holds mdt id
176 struct lu_seq_range {
183 struct lu_seq_range_array {
186 struct lu_seq_range lsra_lsr[0];
189 #define LU_SEQ_RANGE_MDT 0x0
190 #define LU_SEQ_RANGE_OST 0x1
191 #define LU_SEQ_RANGE_ANY 0x3
193 #define LU_SEQ_RANGE_MASK 0x3
195 /** \defgroup lu_fid lu_fid
200 * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat.
201 * Deprecated since HSM and SOM attributes are now stored in separate on-disk
205 LMAC_HSM = 0x00000001,
206 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
207 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
208 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
209 * under /O/<seq>/d<x>.
214 * Masks for all features that should be supported by a Lustre version to
215 * access a specific file.
216 * This information is stored in lustre_mdt_attrs::lma_incompat.
219 LMAI_RELEASED = 0x00000001, /* file is released */
220 LMAI_AGENT = 0x00000002, /* agent inode */
221 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
222 * is on the remote MDT
226 #define LMA_INCOMPAT_SUPP (LMAI_AGENT | LMAI_REMOTE_PARENT)
232 /** LASTID file has zero OID */
233 LUSTRE_FID_LASTID_OID = 0UL,
234 /** initial fid id value */
235 LUSTRE_FID_INIT_OID = 1UL
238 /** returns fid object sequence */
239 static inline __u64 fid_seq(const struct lu_fid *fid)
244 /** returns fid object id */
245 static inline __u32 fid_oid(const struct lu_fid *fid)
250 /** returns fid object version */
251 static inline __u32 fid_ver(const struct lu_fid *fid)
256 static inline void fid_zero(struct lu_fid *fid)
258 memset(fid, 0, sizeof(*fid));
261 static inline __u64 fid_ver_oid(const struct lu_fid *fid)
263 return ((__u64)fid_ver(fid) << 32 | fid_oid(fid));
266 /* copytool uses a 32b bitmask field to encode archive-Ids during register
268 * archive num = 0 => all
269 * archive num from 1 to 32
271 #define LL_HSM_MAX_ARCHIVE (sizeof(__u32) * 8)
274 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
275 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
276 * used for other purposes and not risk collisions with existing inodes.
278 * Different FID Format
279 * http://wiki.old.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs
282 FID_SEQ_OST_MDT0 = 0,
283 FID_SEQ_LLOG = 1, /* unnamed llogs */
285 FID_SEQ_OST_MDT1 = 3,
286 FID_SEQ_OST_MAX = 9, /* Max MDT count before OST_on_FID */
287 FID_SEQ_LLOG_NAME = 10, /* named llogs */
290 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
291 FID_SEQ_IDIF = 0x100000000ULL,
292 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
293 /* Normal FID sequence starts from this value, i.e. 1<<33 */
294 FID_SEQ_START = 0x200000000ULL,
295 /* sequence for local pre-defined FIDs listed in local_oid */
296 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
297 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
298 /* sequence is used for local named objects FIDs generated
299 * by local_object_storage library
301 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
302 /* Because current FLD will only cache the fid sequence, instead
303 * of oid on the client side, if the FID needs to be exposed to
304 * clients sides, it needs to make sure all of fids under one
305 * sequence will be located in one MDT.
307 FID_SEQ_SPECIAL = 0x200000004ULL,
308 FID_SEQ_QUOTA = 0x200000005ULL,
309 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
310 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
311 FID_SEQ_NORMAL = 0x200000400ULL,
312 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
315 #define OBIF_OID_MAX_BITS 32
316 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
317 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
318 #define IDIF_OID_MAX_BITS 48
319 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
320 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
322 /** OID for FID_SEQ_SPECIAL */
324 /* Big Filesystem Lock to serialize rename operations */
325 FID_OID_SPECIAL_BFL = 1UL,
328 /** OID for FID_SEQ_DOT_LUSTRE */
329 enum dot_lustre_oid {
330 FID_OID_DOT_LUSTRE = 1UL,
331 FID_OID_DOT_LUSTRE_OBF = 2UL,
334 static inline bool fid_seq_is_mdt0(__u64 seq)
336 return (seq == FID_SEQ_OST_MDT0);
339 static inline bool fid_seq_is_mdt(__u64 seq)
341 return seq == FID_SEQ_OST_MDT0 || seq >= FID_SEQ_NORMAL;
344 static inline bool fid_seq_is_echo(__u64 seq)
346 return (seq == FID_SEQ_ECHO);
349 static inline bool fid_is_echo(const struct lu_fid *fid)
351 return fid_seq_is_echo(fid_seq(fid));
354 static inline bool fid_seq_is_llog(__u64 seq)
356 return (seq == FID_SEQ_LLOG);
359 static inline bool fid_is_llog(const struct lu_fid *fid)
361 /* file with OID == 0 is not llog but contains last oid */
362 return fid_seq_is_llog(fid_seq(fid)) && fid_oid(fid) > 0;
365 static inline bool fid_seq_is_rsvd(__u64 seq)
367 return (seq > FID_SEQ_OST_MDT0 && seq <= FID_SEQ_RSVD);
370 static inline bool fid_seq_is_special(__u64 seq)
372 return seq == FID_SEQ_SPECIAL;
375 static inline bool fid_seq_is_local_file(__u64 seq)
377 return seq == FID_SEQ_LOCAL_FILE ||
378 seq == FID_SEQ_LOCAL_NAME;
381 static inline bool fid_seq_is_root(__u64 seq)
383 return seq == FID_SEQ_ROOT;
386 static inline bool fid_seq_is_dot(__u64 seq)
388 return seq == FID_SEQ_DOT_LUSTRE;
391 static inline bool fid_seq_is_default(__u64 seq)
393 return seq == FID_SEQ_LOV_DEFAULT;
396 static inline bool fid_is_mdt0(const struct lu_fid *fid)
398 return fid_seq_is_mdt0(fid_seq(fid));
401 static inline void lu_root_fid(struct lu_fid *fid)
403 fid->f_seq = FID_SEQ_ROOT;
409 * Check if a fid is igif or not.
410 * \param fid the fid to be tested.
411 * \return true if the fid is a igif; otherwise false.
413 static inline bool fid_seq_is_igif(__u64 seq)
415 return seq >= FID_SEQ_IGIF && seq <= FID_SEQ_IGIF_MAX;
418 static inline bool fid_is_igif(const struct lu_fid *fid)
420 return fid_seq_is_igif(fid_seq(fid));
424 * Check if a fid is idif or not.
425 * \param fid the fid to be tested.
426 * \return true if the fid is a idif; otherwise false.
428 static inline bool fid_seq_is_idif(__u64 seq)
430 return seq >= FID_SEQ_IDIF && seq <= FID_SEQ_IDIF_MAX;
433 static inline bool fid_is_idif(const struct lu_fid *fid)
435 return fid_seq_is_idif(fid_seq(fid));
438 static inline bool fid_is_local_file(const struct lu_fid *fid)
440 return fid_seq_is_local_file(fid_seq(fid));
443 static inline bool fid_seq_is_norm(__u64 seq)
445 return (seq >= FID_SEQ_NORMAL);
448 static inline bool fid_is_norm(const struct lu_fid *fid)
450 return fid_seq_is_norm(fid_seq(fid));
453 /* convert an OST objid into an IDIF FID SEQ number */
454 static inline __u64 fid_idif_seq(__u64 id, __u32 ost_idx)
456 return FID_SEQ_IDIF | (ost_idx << 16) | ((id >> 32) & 0xffff);
459 /* convert a packed IDIF FID into an OST objid */
460 static inline __u64 fid_idif_id(__u64 seq, __u32 oid, __u32 ver)
462 return ((__u64)ver << 48) | ((seq & 0xffff) << 32) | oid;
465 /* extract ost index from IDIF FID */
466 static inline __u32 fid_idif_ost_idx(const struct lu_fid *fid)
468 return (fid_seq(fid) >> 16) & 0xffff;
471 /* extract OST sequence (group) from a wire ost_id (id/seq) pair */
472 static inline __u64 ostid_seq(const struct ost_id *ostid)
474 if (fid_seq_is_mdt0(ostid->oi.oi_seq))
475 return FID_SEQ_OST_MDT0;
477 if (unlikely(fid_seq_is_default(ostid->oi.oi_seq)))
478 return FID_SEQ_LOV_DEFAULT;
480 if (fid_is_idif(&ostid->oi_fid))
481 return FID_SEQ_OST_MDT0;
483 return fid_seq(&ostid->oi_fid);
486 /* extract OST objid from a wire ost_id (id/seq) pair */
487 static inline __u64 ostid_id(const struct ost_id *ostid)
489 if (fid_seq_is_mdt0(ostid->oi.oi_seq))
490 return ostid->oi.oi_id & IDIF_OID_MASK;
492 if (unlikely(fid_seq_is_default(ostid->oi.oi_seq)))
493 return ostid->oi.oi_id;
495 if (fid_is_idif(&ostid->oi_fid))
496 return fid_idif_id(fid_seq(&ostid->oi_fid),
497 fid_oid(&ostid->oi_fid), 0);
499 return fid_oid(&ostid->oi_fid);
502 static inline void ostid_set_seq(struct ost_id *oi, __u64 seq)
504 if (fid_seq_is_mdt0(seq) || fid_seq_is_default(seq)) {
507 oi->oi_fid.f_seq = seq;
508 /* Note: if f_oid + f_ver is zero, we need init it
509 * to be 1, otherwise, ostid_seq will treat this
510 * as old ostid (oi_seq == 0)
512 if (oi->oi_fid.f_oid == 0 && oi->oi_fid.f_ver == 0)
513 oi->oi_fid.f_oid = LUSTRE_FID_INIT_OID;
517 static inline void ostid_set_seq_mdt0(struct ost_id *oi)
519 ostid_set_seq(oi, FID_SEQ_OST_MDT0);
522 static inline void ostid_set_seq_echo(struct ost_id *oi)
524 ostid_set_seq(oi, FID_SEQ_ECHO);
527 static inline void ostid_set_seq_llog(struct ost_id *oi)
529 ostid_set_seq(oi, FID_SEQ_LLOG);
533 * Note: we need check oi_seq to decide where to set oi_id,
534 * so oi_seq should always be set ahead of oi_id.
536 static inline void ostid_set_id(struct ost_id *oi, __u64 oid)
538 if (fid_seq_is_mdt0(oi->oi.oi_seq)) {
539 if (oid >= IDIF_MAX_OID) {
540 CERROR("Too large OID %#llx to set MDT0 " DOSTID "\n",
545 } else if (fid_is_idif(&oi->oi_fid)) {
546 if (oid >= IDIF_MAX_OID) {
547 CERROR("Too large OID %#llx to set IDIF " DOSTID "\n",
551 oi->oi_fid.f_seq = fid_idif_seq(oid,
552 fid_idif_ost_idx(&oi->oi_fid));
553 oi->oi_fid.f_oid = oid;
554 oi->oi_fid.f_ver = oid >> 48;
556 if (oid >= OBIF_MAX_OID) {
557 CERROR("Bad %llu to set " DOSTID "\n", oid, POSTID(oi));
560 oi->oi_fid.f_oid = oid;
564 static inline int fid_set_id(struct lu_fid *fid, __u64 oid)
566 if (unlikely(fid_seq_is_igif(fid->f_seq))) {
567 CERROR("bad IGIF, "DFID"\n", PFID(fid));
571 if (fid_is_idif(fid)) {
572 if (oid >= IDIF_MAX_OID) {
573 CERROR("Too large OID %#llx to set IDIF " DFID "\n",
574 (unsigned long long)oid, PFID(fid));
577 fid->f_seq = fid_idif_seq(oid, fid_idif_ost_idx(fid));
579 fid->f_ver = oid >> 48;
581 if (oid >= OBIF_MAX_OID) {
582 CERROR("Too large OID %#llx to set REG " DFID "\n",
583 (unsigned long long)oid, PFID(fid));
592 * Unpack an OST object id/seq (group) into a FID. This is needed for
593 * converting all obdo, lmm, lsm, etc. 64-bit id/seq pairs into proper
594 * FIDs. Note that if an id/seq is already in FID/IDIF format it will
595 * be passed through unchanged. Only legacy OST objects in "group 0"
596 * will be mapped into the IDIF namespace so that they can fit into the
597 * struct lu_fid fields without loss. For reference see:
598 * http://wiki.old.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs
600 static inline int ostid_to_fid(struct lu_fid *fid, struct ost_id *ostid,
603 __u64 seq = ostid_seq(ostid);
605 if (ost_idx > 0xffff) {
606 CERROR("bad ost_idx, "DOSTID" ost_idx:%u\n", POSTID(ostid),
611 if (fid_seq_is_mdt0(seq)) {
612 __u64 oid = ostid_id(ostid);
614 /* This is a "legacy" (old 1.x/2.early) OST object in "group 0"
615 * that we map into the IDIF namespace. It allows up to 2^48
616 * objects per OST, as this is the object namespace that has
617 * been in production for years. This can handle create rates
618 * of 1M objects/s/OST for 9 years, or combinations thereof.
620 if (oid >= IDIF_MAX_OID) {
621 CERROR("bad MDT0 id, " DOSTID " ost_idx:%u\n",
622 POSTID(ostid), ost_idx);
625 fid->f_seq = fid_idif_seq(oid, ost_idx);
626 /* truncate to 32 bits by assignment */
628 /* in theory, not currently used */
629 fid->f_ver = oid >> 48;
630 } else if (likely(!fid_seq_is_default(seq))) {
631 /* This is either an IDIF object, which identifies objects across
632 * all OSTs, or a regular FID. The IDIF namespace maps legacy
633 * OST objects into the FID namespace. In both cases, we just
634 * pass the FID through, no conversion needed.
636 if (ostid->oi_fid.f_ver != 0) {
637 CERROR("bad MDT0 id, " DOSTID " ost_idx:%u\n",
638 POSTID(ostid), ost_idx);
641 *fid = ostid->oi_fid;
647 /* pack any OST FID into an ostid (id/seq) for the wire/disk */
648 static inline int fid_to_ostid(const struct lu_fid *fid, struct ost_id *ostid)
650 if (unlikely(fid_seq_is_igif(fid->f_seq))) {
651 CERROR("bad IGIF, "DFID"\n", PFID(fid));
655 if (fid_is_idif(fid)) {
656 ostid_set_seq_mdt0(ostid);
657 ostid_set_id(ostid, fid_idif_id(fid_seq(fid), fid_oid(fid),
660 ostid->oi_fid = *fid;
666 /* Check whether the fid is for LAST_ID */
667 static inline bool fid_is_last_id(const struct lu_fid *fid)
669 return (fid_oid(fid) == 0);
673 * Get inode number from a igif.
674 * \param fid a igif to get inode number from.
675 * \return inode number for the igif.
677 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
682 void lustre_swab_ost_id(struct ost_id *oid);
685 * Get inode generation from a igif.
686 * \param fid a igif to get inode generation from.
687 * \return inode generation for the igif.
689 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
695 * Build igif from the inode number/generation.
697 static inline void lu_igif_build(struct lu_fid *fid, __u32 ino, __u32 gen)
705 * Fids are transmitted across network (in the sender byte-ordering),
706 * and stored on disk in big-endian order.
708 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
710 dst->f_seq = cpu_to_le64(fid_seq(src));
711 dst->f_oid = cpu_to_le32(fid_oid(src));
712 dst->f_ver = cpu_to_le32(fid_ver(src));
715 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
717 dst->f_seq = le64_to_cpu(fid_seq(src));
718 dst->f_oid = le32_to_cpu(fid_oid(src));
719 dst->f_ver = le32_to_cpu(fid_ver(src));
722 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
724 dst->f_seq = cpu_to_be64(fid_seq(src));
725 dst->f_oid = cpu_to_be32(fid_oid(src));
726 dst->f_ver = cpu_to_be32(fid_ver(src));
729 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
731 dst->f_seq = be64_to_cpu(fid_seq(src));
732 dst->f_oid = be32_to_cpu(fid_oid(src));
733 dst->f_ver = be32_to_cpu(fid_ver(src));
736 static inline bool fid_is_sane(const struct lu_fid *fid)
739 ((fid_seq(fid) >= FID_SEQ_START && fid_ver(fid) == 0) ||
740 fid_is_igif(fid) || fid_is_idif(fid) ||
741 fid_seq_is_rsvd(fid_seq(fid)));
744 void lustre_swab_lu_fid(struct lu_fid *fid);
746 static inline bool lu_fid_eq(const struct lu_fid *f0, const struct lu_fid *f1)
748 return memcmp(f0, f1, sizeof(*f0)) == 0;
751 #define __diff_normalize(val0, val1) \
753 typeof(val0) __val0 = (val0); \
754 typeof(val1) __val1 = (val1); \
756 (__val0 == __val1 ? 0 : __val0 > __val1 ? 1 : -1); \
759 static inline int lu_fid_cmp(const struct lu_fid *f0,
760 const struct lu_fid *f1)
763 __diff_normalize(fid_seq(f0), fid_seq(f1)) ?:
764 __diff_normalize(fid_oid(f0), fid_oid(f1)) ?:
765 __diff_normalize(fid_ver(f0), fid_ver(f1));
768 static inline void ostid_cpu_to_le(const struct ost_id *src_oi,
769 struct ost_id *dst_oi)
771 if (fid_seq_is_mdt0(ostid_seq(src_oi))) {
772 dst_oi->oi.oi_id = cpu_to_le64(src_oi->oi.oi_id);
773 dst_oi->oi.oi_seq = cpu_to_le64(src_oi->oi.oi_seq);
775 fid_cpu_to_le(&dst_oi->oi_fid, &src_oi->oi_fid);
779 static inline void ostid_le_to_cpu(const struct ost_id *src_oi,
780 struct ost_id *dst_oi)
782 if (fid_seq_is_mdt0(ostid_seq(src_oi))) {
783 dst_oi->oi.oi_id = le64_to_cpu(src_oi->oi.oi_id);
784 dst_oi->oi.oi_seq = le64_to_cpu(src_oi->oi.oi_seq);
786 fid_le_to_cpu(&dst_oi->oi_fid, &src_oi->oi_fid);
792 /** \defgroup lu_dir lu_dir
797 * Enumeration of possible directory entry attributes.
799 * Attributes follow directory entry header in the order they appear in this
802 enum lu_dirent_attrs {
805 LUDA_64BITHASH = 0x0004,
809 * Layout of readdir pages, as transmitted on wire.
812 /** valid if LUDA_FID is set. */
813 struct lu_fid lde_fid;
814 /** a unique entry identifier: a hash or an offset. */
816 /** total record length, including all attributes. */
820 /** optional variable size attributes following this entry.
821 * taken from enum lu_dirent_attrs.
824 /** name is followed by the attributes indicated in ->ldp_attrs, in
825 * their natural order. After the last attribute, padding bytes are
826 * added to make ->lde_reclen a multiple of 8.
832 * Definitions of optional directory entry attributes formats.
834 * Individual attributes do not have their length encoded in a generic way. It
835 * is assumed that consumer of an attribute knows its format. This means that
836 * it is impossible to skip over an unknown attribute, except by skipping over all
837 * remaining attributes (by using ->lde_reclen), which is not too
838 * constraining, because new server versions will append new attributes at
839 * the end of an entry.
843 * Fid directory attribute: a fid of an object referenced by the entry. This
844 * will be almost always requested by the client and supplied by the server.
846 * Aligned to 8 bytes.
848 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
853 * Aligned to 2 bytes.
864 #define IFTODT(type) (((type) & S_IFMT) >> IFSHIFT)
867 #define DTTOIF(dirtype) ((dirtype) << IFSHIFT)
871 __u64 ldp_hash_start;
875 struct lu_dirent ldp_entries[0];
878 enum lu_dirpage_flags {
880 * dirpage contains no entry.
884 * last entry's lde_hash equals ldp_hash_end.
889 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
891 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
894 return dp->ldp_entries;
897 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
899 struct lu_dirent *next;
901 if (le16_to_cpu(ent->lde_reclen) != 0)
902 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
909 static inline size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
913 if (attr & LUDA_TYPE) {
914 const size_t align = sizeof(struct luda_type) - 1;
916 size = (sizeof(struct lu_dirent) + namelen + align) & ~align;
917 size += sizeof(struct luda_type);
919 size = sizeof(struct lu_dirent) + namelen;
922 return (size + 7) & ~7;
925 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
928 * MDS_READPAGE page size
930 * This is the directory page size packed in MDS_READPAGE RPC.
931 * It's different than PAGE_SIZE because the client needs to
932 * access the struct lu_dirpage header packed at the beginning of
933 * the "page" and without this there isn't any way to know find the
934 * lu_dirpage header is if client and server PAGE_SIZE differ.
936 #define LU_PAGE_SHIFT 12
937 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
938 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
940 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
944 struct lustre_handle {
948 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
950 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
952 return lh->cookie != 0ull;
955 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
956 const struct lustre_handle *lh2)
958 return lh1->cookie == lh2->cookie;
961 static inline void lustre_handle_copy(struct lustre_handle *tgt,
962 const struct lustre_handle *src)
964 tgt->cookie = src->cookie;
967 /* flags for lm_flags */
968 #define MSGHDR_AT_SUPPORT 0x1
969 #define MSGHDR_CKSUM_INCOMPAT18 0x2
971 #define lustre_msg lustre_msg_v2
972 /* we depend on this structure to be 8-byte aligned */
973 /* this type is only endian-adjusted in lustre_unpack_msg() */
974 struct lustre_msg_v2 {
986 /* without gss, ptlrpc_body is put at the first buffer. */
987 #define PTLRPC_NUM_VERSIONS 4
989 struct ptlrpc_body_v3 {
990 struct lustre_handle pb_handle;
995 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
996 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
999 __u64 pb_last_committed;
1004 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1005 __u32 pb_service_time; /* for rep, actual service time */
1008 /* VBR: pre-versions */
1009 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1010 __u64 pb_mbits; /**< match bits for bulk request */
1011 /* padding for future needs */
1012 __u64 pb_padding64_0;
1013 __u64 pb_padding64_1;
1014 __u64 pb_padding64_2;
1015 char pb_jobid[LUSTRE_JOBID_SIZE];
1018 #define ptlrpc_body ptlrpc_body_v3
1020 struct ptlrpc_body_v2 {
1021 struct lustre_handle pb_handle;
1026 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
1027 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
1030 __u64 pb_last_committed;
1035 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1036 __u32 pb_service_time; /* for rep, actual service time, also used for
1037 * net_latency of req
1041 /* VBR: pre-versions */
1042 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1043 __u64 pb_mbits; /**< unused in V2 */
1044 /* padding for future needs */
1045 __u64 pb_padding64_0;
1046 __u64 pb_padding64_1;
1047 __u64 pb_padding64_2;
1050 void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
1052 /* message body offset for lustre_msg_v2 */
1053 /* ptlrpc body offset in all request/reply messages */
1054 #define MSG_PTLRPC_BODY_OFF 0
1056 /* normal request/reply message record offset */
1057 #define REQ_REC_OFF 1
1058 #define REPLY_REC_OFF 1
1060 /* ldlm request message body offset */
1061 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
1062 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
1064 /* ldlm intent lock message body offset */
1065 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
1066 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
1068 /* ldlm reply message body offset */
1069 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
1070 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
1072 /** only use in req->rq_{req,rep}_swab_mask */
1073 #define MSG_PTLRPC_HEADER_OFF 31
1075 /* Flags that are operation-specific go in the top 16 bits. */
1076 #define MSG_OP_FLAG_MASK 0xffff0000
1077 #define MSG_OP_FLAG_SHIFT 16
1079 /* Flags that apply to all requests are in the bottom 16 bits */
1080 #define MSG_GEN_FLAG_MASK 0x0000ffff
1081 #define MSG_LAST_REPLAY 0x0001
1082 #define MSG_RESENT 0x0002
1083 #define MSG_REPLAY 0x0004
1084 /* #define MSG_AT_SUPPORT 0x0008
1085 * This was used in early prototypes of adaptive timeouts, and while there
1086 * shouldn't be any users of that code there also isn't a need for using this
1087 * bits. Defer usage until at least 1.10 to avoid potential conflict.
1089 #define MSG_DELAY_REPLAY 0x0010
1090 #define MSG_VERSION_REPLAY 0x0020
1091 #define MSG_REQ_REPLAY_DONE 0x0040
1092 #define MSG_LOCK_REPLAY_DONE 0x0080
1095 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
1098 #define MSG_CONNECT_RECOVERING 0x00000001
1099 #define MSG_CONNECT_RECONNECT 0x00000002
1100 #define MSG_CONNECT_REPLAYABLE 0x00000004
1101 /*#define MSG_CONNECT_PEER 0x8 */
1102 #define MSG_CONNECT_LIBCLIENT 0x00000010
1103 #define MSG_CONNECT_INITIAL 0x00000020
1104 #define MSG_CONNECT_ASYNC 0x00000040
1105 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
1106 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
1109 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
1110 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
1111 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
1112 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
1113 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
1114 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
1115 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
1116 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
1117 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
1118 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
1119 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
1120 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
1121 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
1122 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
1123 *We do not support JOIN FILE
1124 *anymore, reserve this flags
1125 *just for preventing such bit
1128 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
1129 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
1130 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
1131 * in production. Removed in
1132 * 2.9. Keep this flag to
1135 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
1136 * never used in production.
1137 * Removed in 2.9. Keep this
1138 * flag to avoid reuse
1140 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
1141 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
1142 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
1143 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
1144 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
1145 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
1146 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
1147 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
1148 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
1149 #define OBD_CONNECT_REAL 0x8000000ULL /*real connection */
1150 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
1151 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
1152 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
1153 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
1154 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
1155 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
1156 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
1157 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
1158 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
1159 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
1160 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
1163 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
1164 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
1165 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
1166 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
1167 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
1168 * RPC error properly
1170 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
1171 * finer space reservation
1173 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
1174 * policy and 2.x server
1176 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
1177 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
1178 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
1179 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
1180 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
1181 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* flock deadlock detection */
1182 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/*create stripe disposition*/
1183 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
1186 #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */
1187 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */
1188 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify
1191 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL/* striped DNE dir */
1192 /** bulk matchbits is sent within ptlrpc_body */
1193 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL
1196 * Please DO NOT add flag values here before first ensuring that this same
1197 * flag value is not in use on some other branch. Please clear any such
1198 * changes with senior engineers before starting to use a new flag. Then,
1199 * submit a small patch against EVERY branch that ONLY adds the new flag,
1200 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
1201 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
1202 * can be approved and landed easily to reserve the flag for future use.
1205 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
1206 * connection. It is a temporary bug fix for Imperative Recovery interop
1207 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
1208 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644.
1210 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
1212 #define OCD_HAS_FLAG(ocd, flg) \
1213 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
1215 /* Features required for this version of the client to work with server */
1216 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
1219 /* This structure is used for both request and reply.
1221 * If we eventually have separate connect data for different types, which we
1222 * almost certainly will, then perhaps we stick a union in here.
1224 struct obd_connect_data {
1225 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
1226 __u32 ocd_version; /* lustre release version number */
1227 __u32 ocd_grant; /* initial cache grant amount (bytes) */
1228 __u32 ocd_index; /* LOV index to connect to */
1229 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
1230 __u64 ocd_ibits_known; /* inode bits this client understands */
1231 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
1232 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
1233 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
1234 __u32 ocd_unused; /* also fix lustre_swab_connect */
1235 __u64 ocd_transno; /* first transno from client to be replayed */
1236 __u32 ocd_group; /* MDS group on OST */
1237 __u32 ocd_cksum_types; /* supported checksum algorithms */
1238 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1239 __u32 ocd_instance; /* instance # of this target */
1240 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1241 /* Fields after ocd_maxbytes are only accessible by the receiver
1242 * if the corresponding flag in ocd_connect_flags is set. Accessing
1243 * any field after ocd_maxbytes on the receiver without a valid flag
1244 * may result in out-of-bound memory access and kernel oops.
1246 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
1247 __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */
1248 __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */
1249 __u64 padding2; /* added 2.1.0. also fix lustre_swab_connect */
1250 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
1251 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
1252 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
1253 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
1254 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
1255 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
1256 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
1257 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
1258 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
1259 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
1260 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
1261 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
1262 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1266 * Please DO NOT use any fields here before first ensuring that this same
1267 * field is not in use on some other branch. Please clear any such changes
1268 * with senior engineers before starting to use a new field. Then, submit
1269 * a small patch against EVERY branch that ONLY adds the new field along with
1270 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
1271 * reserve the flag for future use.
1274 void lustre_swab_connect(struct obd_connect_data *ocd);
1277 * Supported checksum algorithms. Up to 32 checksum types are supported.
1278 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1279 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
1280 * algorithm and also the OBD_FL_CKSUM* flags.
1283 OBD_CKSUM_CRC32 = 0x00000001,
1284 OBD_CKSUM_ADLER = 0x00000002,
1285 OBD_CKSUM_CRC32C = 0x00000004,
1289 * OST requests: OBDO & OBD request records
1294 OST_REPLY = 0, /* reply ? */
1310 OST_QUOTACHECK = 18, /* not used since 2.4 */
1312 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1315 #define OST_FIRST_OPC OST_REPLY
1318 OBD_FL_INLINEDATA = 0x00000001,
1319 OBD_FL_OBDMDEXISTS = 0x00000002,
1320 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1321 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1322 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1323 OBD_FL_RECREATE_OBJS = 0x00000020, /* recreate missing obj */
1324 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1325 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1326 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1327 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1328 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1329 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1330 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1331 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1332 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
1333 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1334 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1335 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1336 * XXX: obsoleted - reserved for old
1337 * clients prior than 2.2
1339 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1340 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1341 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1342 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1344 /* Note that while these checksum values are currently separate bits,
1345 * in 2.x we can actually allow all values from 1-31 if we wanted.
1347 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1348 OBD_FL_CKSUM_CRC32C,
1350 /* mask for local-only flag, which won't be sent over network */
1351 OBD_FL_LOCAL_MASK = 0xF0000000,
1355 * All LOV EA magics should have the same postfix, if some new version
1356 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1357 * Lustre, even though the old version system does not recognizes such
1358 * new magic, it still can distinguish the corrupted cases by checking
1359 * the magic's postfix.
1361 #define LOV_MAGIC_MAGIC 0x0BD0
1362 #define LOV_MAGIC_MASK 0xFFFF
1364 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1365 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1366 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1367 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1368 /* reserved for specifying OSTs */
1369 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1370 #define LOV_MAGIC LOV_MAGIC_V1
1373 * magic for fully defined striping
1374 * the idea is that we should have different magics for striping "hints"
1375 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1376 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1377 * we can't just change it w/o long way preparation, but we still need a
1378 * mechanism to allow LOD to differentiate hint versus ready striping.
1379 * so, at the moment we do a trick: MDT knows what to expect from request
1380 * depending on the case (replay uses ready striping, non-replay req uses
1381 * hints), so MDT replaces magic with appropriate one and now LOD can
1382 * easily understand what's inside -bzzz
1384 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
1385 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
1387 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1388 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1390 #define lov_ost_data lov_ost_data_v1
1391 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1392 struct ost_id l_ost_oi; /* OST object ID */
1393 __u32 l_ost_gen; /* generation of this l_ost_idx */
1394 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1397 #define lov_mds_md lov_mds_md_v1
1398 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1399 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1400 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1401 struct ost_id lmm_oi; /* LOV object ID */
1402 __u32 lmm_stripe_size; /* size of stripe in bytes */
1403 /* lmm_stripe_count used to be __u32 */
1404 __u16 lmm_stripe_count; /* num stripes in use for this object */
1405 __u16 lmm_layout_gen; /* layout generation number */
1406 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1410 * Sigh, because pre-2.4 uses
1411 * struct lov_mds_md_v1 {
1413 * __u64 lmm_object_id;
1414 * __u64 lmm_object_seq;
1417 * to identify the LOV(MDT) object, and lmm_object_seq will
1418 * be normal_fid, which make it hard to combine these conversion
1419 * to ostid_to FID. so we will do lmm_oi/fid conversion separately
1421 * We can tell the lmm_oi by this way,
1422 * 1.8: lmm_object_id = {inode}, lmm_object_gr = 0
1423 * 2.1: lmm_object_id = {oid < 128k}, lmm_object_seq = FID_SEQ_NORMAL
1424 * 2.4: lmm_oi.f_seq = FID_SEQ_NORMAL, lmm_oi.f_oid = {oid < 128k},
1427 * But currently lmm_oi/lsm_oi does not have any "real" usages,
1428 * except for printing some information, and the user can always
1429 * get the real FID from LMA, besides this multiple case check might
1430 * make swab more complicate. So we will keep using id/seq for lmm_oi.
1433 static inline void fid_to_lmm_oi(const struct lu_fid *fid,
1436 oi->oi.oi_id = fid_oid(fid);
1437 oi->oi.oi_seq = fid_seq(fid);
1440 static inline void lmm_oi_set_seq(struct ost_id *oi, __u64 seq)
1442 oi->oi.oi_seq = seq;
1445 static inline void lmm_oi_set_id(struct ost_id *oi, __u64 oid)
1450 static inline __u64 lmm_oi_id(const struct ost_id *oi)
1452 return oi->oi.oi_id;
1455 static inline __u64 lmm_oi_seq(const struct ost_id *oi)
1457 return oi->oi.oi_seq;
1460 static inline void lmm_oi_le_to_cpu(struct ost_id *dst_oi,
1461 const struct ost_id *src_oi)
1463 dst_oi->oi.oi_id = le64_to_cpu(src_oi->oi.oi_id);
1464 dst_oi->oi.oi_seq = le64_to_cpu(src_oi->oi.oi_seq);
1467 static inline void lmm_oi_cpu_to_le(struct ost_id *dst_oi,
1468 const struct ost_id *src_oi)
1470 dst_oi->oi.oi_id = cpu_to_le64(src_oi->oi.oi_id);
1471 dst_oi->oi.oi_seq = cpu_to_le64(src_oi->oi.oi_seq);
1474 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
1476 #define MAX_MD_SIZE \
1477 (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1478 #define MIN_MD_SIZE \
1479 (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1481 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1482 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1483 #define XATTR_USER_PREFIX "user."
1484 #define XATTR_TRUSTED_PREFIX "trusted."
1485 #define XATTR_SECURITY_PREFIX "security."
1486 #define XATTR_LUSTRE_PREFIX "lustre."
1488 #define XATTR_NAME_LOV "trusted.lov"
1489 #define XATTR_NAME_LMA "trusted.lma"
1490 #define XATTR_NAME_LMV "trusted.lmv"
1491 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1492 #define XATTR_NAME_LINK "trusted.link"
1493 #define XATTR_NAME_FID "trusted.fid"
1494 #define XATTR_NAME_VERSION "trusted.version"
1495 #define XATTR_NAME_SOM "trusted.som"
1496 #define XATTR_NAME_HSM "trusted.hsm"
1497 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_namespace"
1499 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1500 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1501 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1502 struct ost_id lmm_oi; /* LOV object ID */
1503 __u32 lmm_stripe_size; /* size of stripe in bytes */
1504 /* lmm_stripe_count used to be __u32 */
1505 __u16 lmm_stripe_count; /* num stripes in use for this object */
1506 __u16 lmm_layout_gen; /* layout generation number */
1507 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1508 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1511 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1513 if (lmm_magic == LOV_MAGIC_V3)
1514 return sizeof(struct lov_mds_md_v3) +
1515 stripes * sizeof(struct lov_ost_data_v1);
1517 return sizeof(struct lov_mds_md_v1) +
1518 stripes * sizeof(struct lov_ost_data_v1);
1522 lov_mds_md_max_stripe_count(size_t buf_size, __u32 lmm_magic)
1524 switch (lmm_magic) {
1525 case LOV_MAGIC_V1: {
1526 struct lov_mds_md_v1 lmm;
1528 if (buf_size < sizeof(lmm))
1531 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1533 case LOV_MAGIC_V3: {
1534 struct lov_mds_md_v3 lmm;
1536 if (buf_size < sizeof(lmm))
1539 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1546 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1547 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1548 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1549 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1550 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1551 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1552 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1553 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1554 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1555 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1556 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1557 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1558 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1559 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1560 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
1561 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1562 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1563 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1564 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1565 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1566 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1567 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1568 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1569 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1570 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1571 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1572 /* ->mds if epoch opens or closes
1574 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1575 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1576 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1577 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1578 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1580 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1581 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1582 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1583 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1585 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1586 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1587 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1588 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1589 /* OBD_MD_FLRMTPERM (0x0000010000000000ULL) remote perm, obsolete */
1590 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1591 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1592 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1593 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1594 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1595 * under lock; for xattr
1596 * requests means the
1597 * client holds the lock
1599 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1601 /* OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) lfs lsetfacl, obsolete */
1602 /* OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) lfs lgetfacl, obsolete */
1603 /* OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) lfs rsetfacl, obsolete */
1604 /* OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) lfs rgetfacl, obsolete */
1606 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1607 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1611 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1613 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1614 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1615 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1616 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1617 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1619 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1621 /* don't forget obdo_fid which is way down at the bottom so it can
1622 * come after the definition of llog_cookie
1627 HSS_CLEARMASK = 0x02,
1628 HSS_ARCHIVE_ID = 0x04,
1631 struct hsm_state_set {
1633 __u32 hss_archive_id;
1635 __u64 hss_clearmask;
1638 void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
1639 void lustre_swab_hsm_state_set(struct hsm_state_set *hss);
1641 void lustre_swab_obd_statfs(struct obd_statfs *os);
1643 /* ost_body.data values for OST_BRW */
1645 #define OBD_BRW_READ 0x01
1646 #define OBD_BRW_WRITE 0x02
1647 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1648 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1649 * transfer and is not accounted in
1652 #define OBD_BRW_CHECK 0x10
1653 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1654 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1655 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1656 #define OBD_BRW_NOQUOTA 0x100
1657 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1658 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1659 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1660 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1661 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1662 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1663 * that the client is running low on
1664 * space for unstable pages; asking
1665 * it to sync quickly
1668 #define OBD_OBJECT_EOF LUSTRE_EOF
1670 #define OST_MIN_PRECREATE 32
1671 #define OST_MAX_PRECREATE 20000
1674 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1675 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1676 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1677 * high 16 bits in 2.4 and later
1679 __u32 ioo_bufcnt; /* number of niobufs for this object */
1682 #define IOOBJ_MAX_BRW_BITS 16
1683 #define IOOBJ_TYPE_MASK ((1U << IOOBJ_MAX_BRW_BITS) - 1)
1684 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1685 #define ioobj_max_brw_set(ioo, num) \
1686 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1688 void lustre_swab_obd_ioobj(struct obd_ioobj *ioo);
1690 /* multiple of 8 bytes => can array */
1691 struct niobuf_remote {
1697 void lustre_swab_niobuf_remote(struct niobuf_remote *nbr);
1699 /* lock value block communicated between the filter and llite */
1701 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1702 * negative, i.e. because ((MASK + rc) & MASK) != MASK.
1704 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1705 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1706 #define OST_LVB_IS_ERR(blocks) \
1707 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1708 #define OST_LVB_SET_ERR(blocks, rc) \
1709 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1710 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1720 void lustre_swab_ost_lvb_v1(struct ost_lvb_v1 *lvb);
1734 void lustre_swab_ost_lvb(struct ost_lvb *lvb);
1737 * lquota data structures
1740 /* The lquota_id structure is an union of all the possible identifier types that
1741 * can be used with quota, this includes:
1744 * - a FID which can be used for per-directory quota in the future
1747 struct lu_fid qid_fid; /* FID for per-directory quota */
1748 __u64 qid_uid; /* user identifier */
1749 __u64 qid_gid; /* group identifier */
1752 /* quotactl management */
1753 struct obd_quotactl {
1755 __u32 qc_type; /* see Q_* flag below */
1758 struct obd_dqinfo qc_dqinfo;
1759 struct obd_dqblk qc_dqblk;
1762 void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1764 #define Q_COPY(out, in, member) (out)->member = (in)->member
1766 #define QCTL_COPY(out, in) \
1768 Q_COPY(out, in, qc_cmd); \
1769 Q_COPY(out, in, qc_type); \
1770 Q_COPY(out, in, qc_id); \
1771 Q_COPY(out, in, qc_stat); \
1772 Q_COPY(out, in, qc_dqinfo); \
1773 Q_COPY(out, in, qc_dqblk); \
1776 /* Data structures associated with the quota locks */
1778 /* Glimpse descriptor used for the index & per-ID quota locks */
1779 struct ldlm_gl_lquota_desc {
1780 union lquota_id gl_id; /* quota ID subject to the glimpse */
1781 __u64 gl_flags; /* see LQUOTA_FL* below */
1782 __u64 gl_ver; /* new index version */
1783 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1784 __u64 gl_softlimit; /* new softlimit */
1789 /* quota glimpse flags */
1790 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1792 /* LVB used with quota (global and per-ID) locks */
1794 __u64 lvb_flags; /* see LQUOTA_FL* above */
1795 __u64 lvb_id_may_rel; /* space that might be released later */
1796 __u64 lvb_id_rel; /* space released by the slave for this ID */
1797 __u64 lvb_id_qunit; /* current qunit value */
1801 void lustre_swab_lquota_lvb(struct lquota_lvb *lvb);
1809 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1818 MDS_GETATTR_NAME = 34,
1823 MDS_DISCONNECT = 39,
1826 MDS_PIN = 42, /* obsolete, never used in a release */
1827 MDS_UNPIN = 43, /* obsolete, never used in a release */
1829 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1831 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1834 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1836 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1838 MDS_HSM_STATE_GET = 54,
1839 MDS_HSM_STATE_SET = 55,
1840 MDS_HSM_ACTION = 56,
1841 MDS_HSM_PROGRESS = 57,
1842 MDS_HSM_REQUEST = 58,
1843 MDS_HSM_CT_REGISTER = 59,
1844 MDS_HSM_CT_UNREGISTER = 60,
1845 MDS_SWAP_LAYOUTS = 61,
1849 #define MDS_FIRST_OPC MDS_GETATTR
1855 enum mdt_reint_cmd {
1868 void lustre_swab_generic_32s(__u32 *val);
1870 /* the disposition of the intent outlines what was executed */
1871 #define DISP_IT_EXECD 0x00000001
1872 #define DISP_LOOKUP_EXECD 0x00000002
1873 #define DISP_LOOKUP_NEG 0x00000004
1874 #define DISP_LOOKUP_POS 0x00000008
1875 #define DISP_OPEN_CREATE 0x00000010
1876 #define DISP_OPEN_OPEN 0x00000020
1877 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1878 #define DISP_ENQ_OPEN_REF 0x00800000
1879 #define DISP_ENQ_CREATE_REF 0x01000000
1880 #define DISP_OPEN_LOCK 0x02000000
1881 #define DISP_OPEN_LEASE 0x04000000
1882 #define DISP_OPEN_STRIPE 0x08000000
1883 #define DISP_OPEN_DENY 0x10000000
1885 /* INODE LOCK PARTS */
1886 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
1887 * was used to protect permission (mode,
1888 * owner, group etc) before 2.4.
1890 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1891 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1892 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1894 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
1895 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
1896 * Because for remote directories(in DNE), these locks will be granted by
1897 * different MDTs(different ldlm namespace).
1899 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
1900 * For Remote directory, the master MDT, where the remote directory is, will
1901 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
1902 * will grant LOOKUP_LOCK.
1904 #define MDS_INODELOCK_PERM 0x000010
1905 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
1907 #define MDS_INODELOCK_MAXSHIFT 5
1908 /* This FULL lock is useful to take on unlink sort of operations */
1909 #define MDS_INODELOCK_FULL ((1 << (MDS_INODELOCK_MAXSHIFT + 1)) - 1)
1911 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1912 * but was moved into name[1] along with the OID to avoid consuming the
1913 * name[2,3] fields that need to be used for the quota id (also a FID).
1916 LUSTRE_RES_ID_SEQ_OFF = 0,
1917 LUSTRE_RES_ID_VER_OID_OFF = 1,
1918 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1919 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1920 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1921 LUSTRE_RES_ID_HSH_OFF = 3
1924 #define MDS_STATUS_CONN 1
1925 #define MDS_STATUS_LOV 2
1927 /* these should be identical to their EXT4_*_FL counterparts, they are
1928 * redefined here only to avoid dragging in fs/ext4/ext4.h
1930 #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */
1931 #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */
1932 #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */
1933 #define LUSTRE_NODUMP_FL 0x00000040 /* do not dump file */
1934 #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */
1935 #define LUSTRE_INDEX_FL 0x00001000 /* hash-indexed directory */
1936 #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1937 #define LUSTRE_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
1938 #define LUSTRE_DIRECTIO_FL 0x00100000 /* Use direct i/o */
1939 #define LUSTRE_INLINE_DATA_FL 0x10000000 /* Inode has inline data. */
1941 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1942 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1943 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1944 * the S_* flags are kernel-internal values that change between kernel
1945 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1946 * See b=16526 for a full history.
1948 static inline int ll_ext_to_inode_flags(int flags)
1950 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1951 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1952 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1953 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1954 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0));
1957 static inline int ll_inode_to_ext_flags(int iflags)
1959 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1960 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1961 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1962 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1963 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0));
1966 /* 64 possible states */
1967 enum md_transient_state {
1968 MS_RESTORE = (1 << 0), /* restore is running */
1972 struct lu_fid mbo_fid1;
1973 struct lu_fid mbo_fid2;
1974 struct lustre_handle mbo_handle;
1976 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1980 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1982 __u64 mbo_t_state; /* transient file state defined in
1983 * enum md_transient_state
1984 * was "ino" until 2.4.0
1988 __u32 mbo_capability;
1992 __u32 mbo_flags; /* LUSTRE_*_FL file attributes */
1994 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1995 __u32 mbo_unused2; /* was "generation" until 2.4.0 */
1997 __u32 mbo_eadatasize;
1999 __u32 mbo_max_mdsize;
2000 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
2001 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
2002 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
2003 __u32 mbo_padding_5; /* also fix lustre_swab_mdt_body */
2004 __u64 mbo_padding_6;
2005 __u64 mbo_padding_7;
2006 __u64 mbo_padding_8;
2007 __u64 mbo_padding_9;
2008 __u64 mbo_padding_10;
2011 void lustre_swab_mdt_body(struct mdt_body *b);
2013 struct mdt_ioepoch {
2014 struct lustre_handle mio_handle;
2015 __u64 mio_unused1; /* was ioepoch */
2016 __u32 mio_unused2; /* was flags */
2020 void lustre_swab_mdt_ioepoch(struct mdt_ioepoch *b);
2022 /* permissions for md_perm.mp_perm */
2024 CFS_SETUID_PERM = 0x01,
2025 CFS_SETGID_PERM = 0x02,
2026 CFS_SETGRP_PERM = 0x04,
2029 struct mdt_rec_setattr {
2039 __u32 sa_padding_1_h;
2040 struct lu_fid sa_fid;
2049 __u32 sa_attr_flags;
2051 __u32 sa_bias; /* some operation flags */
2057 void lustre_swab_mdt_rec_setattr(struct mdt_rec_setattr *sa);
2060 * Attribute flags used in mdt_rec_setattr::sa_valid.
2061 * The kernel's #defines for ATTR_* should not be used over the network
2062 * since the client and MDS may run different kernels (see bug 13828)
2063 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
2065 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
2066 #define MDS_ATTR_UID 0x2ULL /* = 2 */
2067 #define MDS_ATTR_GID 0x4ULL /* = 4 */
2068 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
2069 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
2070 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
2071 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
2072 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
2073 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
2074 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
2075 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
2076 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
2077 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
2078 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
2079 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path,
2082 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
2084 #define MDS_FMODE_CLOSED 00000000
2085 #define MDS_FMODE_EXEC 00000004
2086 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
2087 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
2088 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
2090 #define MDS_OPEN_CREATED 00000010
2091 #define MDS_OPEN_CROSS 00000020
2093 #define MDS_OPEN_CREAT 00000100
2094 #define MDS_OPEN_EXCL 00000200
2095 #define MDS_OPEN_TRUNC 00001000
2096 #define MDS_OPEN_APPEND 00002000
2097 #define MDS_OPEN_SYNC 00010000
2098 #define MDS_OPEN_DIRECTORY 00200000
2100 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
2101 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
2102 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
2103 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
2104 * We do not support JOIN FILE
2105 * anymore, reserve this flags
2106 * just for preventing such bit
2110 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
2111 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
2112 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
2113 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
2114 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
2116 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
2118 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
2119 * delegation, succeed if it's not
2120 * being opened with conflict mode.
2122 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
2124 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
2125 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
2126 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
2130 MDS_CHECK_SPLIT = 1 << 0,
2131 MDS_CROSS_REF = 1 << 1,
2132 MDS_VTX_BYPASS = 1 << 2,
2133 MDS_PERM_BYPASS = 1 << 3,
2134 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
2135 MDS_QUOTA_IGNORE = 1 << 5,
2136 MDS_CLOSE_CLEANUP = 1 << 6,
2137 MDS_KEEP_ORPHAN = 1 << 7,
2138 MDS_RECOV_OPEN = 1 << 8,
2139 MDS_DATA_MODIFIED = 1 << 9,
2140 MDS_CREATE_VOLATILE = 1 << 10,
2141 MDS_OWNEROVERRIDE = 1 << 11,
2142 MDS_HSM_RELEASE = 1 << 12,
2143 MDS_RENAME_MIGRATE = BIT(13),
2144 MDS_CLOSE_LAYOUT_SWAP = BIT(14),
2147 /* instance of mdt_reint_rec */
2148 struct mdt_rec_create {
2156 __u32 cr_suppgid1_h;
2158 __u32 cr_suppgid2_h;
2159 struct lu_fid cr_fid1;
2160 struct lu_fid cr_fid2;
2161 struct lustre_handle cr_old_handle; /* handle in case of open replay */
2165 __u64 cr_padding_1; /* rr_blocks */
2168 /* use of helpers set/get_mrc_cr_flags() is needed to access
2169 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2170 * extend cr_flags size without breaking 1.8 compat
2172 __u32 cr_flags_l; /* for use with open, low 32 bits */
2173 __u32 cr_flags_h; /* for use with open, high 32 bits */
2174 __u32 cr_umask; /* umask for create */
2175 __u32 cr_padding_4; /* rr_padding_4 */
2178 static inline void set_mrc_cr_flags(struct mdt_rec_create *mrc, __u64 flags)
2180 mrc->cr_flags_l = (__u32)(flags & 0xFFFFFFFFUll);
2181 mrc->cr_flags_h = (__u32)(flags >> 32);
2184 static inline __u64 get_mrc_cr_flags(struct mdt_rec_create *mrc)
2186 return ((__u64)(mrc->cr_flags_l) | ((__u64)mrc->cr_flags_h << 32));
2189 /* instance of mdt_reint_rec */
2190 struct mdt_rec_link {
2198 __u32 lk_suppgid1_h;
2200 __u32 lk_suppgid2_h;
2201 struct lu_fid lk_fid1;
2202 struct lu_fid lk_fid2;
2204 __u64 lk_padding_1; /* rr_atime */
2205 __u64 lk_padding_2; /* rr_ctime */
2206 __u64 lk_padding_3; /* rr_size */
2207 __u64 lk_padding_4; /* rr_blocks */
2209 __u32 lk_padding_5; /* rr_mode */
2210 __u32 lk_padding_6; /* rr_flags */
2211 __u32 lk_padding_7; /* rr_padding_2 */
2212 __u32 lk_padding_8; /* rr_padding_3 */
2213 __u32 lk_padding_9; /* rr_padding_4 */
2216 /* instance of mdt_reint_rec */
2217 struct mdt_rec_unlink {
2225 __u32 ul_suppgid1_h;
2227 __u32 ul_suppgid2_h;
2228 struct lu_fid ul_fid1;
2229 struct lu_fid ul_fid2;
2231 __u64 ul_padding_2; /* rr_atime */
2232 __u64 ul_padding_3; /* rr_ctime */
2233 __u64 ul_padding_4; /* rr_size */
2234 __u64 ul_padding_5; /* rr_blocks */
2237 __u32 ul_padding_6; /* rr_flags */
2238 __u32 ul_padding_7; /* rr_padding_2 */
2239 __u32 ul_padding_8; /* rr_padding_3 */
2240 __u32 ul_padding_9; /* rr_padding_4 */
2243 /* instance of mdt_reint_rec */
2244 struct mdt_rec_rename {
2252 __u32 rn_suppgid1_h;
2254 __u32 rn_suppgid2_h;
2255 struct lu_fid rn_fid1;
2256 struct lu_fid rn_fid2;
2258 __u64 rn_padding_1; /* rr_atime */
2259 __u64 rn_padding_2; /* rr_ctime */
2260 __u64 rn_padding_3; /* rr_size */
2261 __u64 rn_padding_4; /* rr_blocks */
2262 __u32 rn_bias; /* some operation flags */
2263 __u32 rn_mode; /* cross-ref rename has mode */
2264 __u32 rn_padding_5; /* rr_flags */
2265 __u32 rn_padding_6; /* rr_padding_2 */
2266 __u32 rn_padding_7; /* rr_padding_3 */
2267 __u32 rn_padding_8; /* rr_padding_4 */
2270 /* instance of mdt_reint_rec */
2271 struct mdt_rec_setxattr {
2279 __u32 sx_suppgid1_h;
2281 __u32 sx_suppgid2_h;
2282 struct lu_fid sx_fid;
2283 __u64 sx_padding_1; /* These three are rr_fid2 */
2288 __u64 sx_padding_5; /* rr_ctime */
2289 __u64 sx_padding_6; /* rr_size */
2290 __u64 sx_padding_7; /* rr_blocks */
2293 __u32 sx_padding_8; /* rr_flags */
2294 __u32 sx_padding_9; /* rr_padding_2 */
2295 __u32 sx_padding_10; /* rr_padding_3 */
2296 __u32 sx_padding_11; /* rr_padding_4 */
2300 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2301 * Do NOT change the size of various members, otherwise the value
2302 * will be broken in lustre_swab_mdt_rec_reint().
2304 * If you add new members in other mdt_reint_xxx structures and need to use the
2305 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2307 struct mdt_rec_reint {
2315 __u32 rr_suppgid1_h;
2317 __u32 rr_suppgid2_h;
2318 struct lu_fid rr_fid1;
2319 struct lu_fid rr_fid2;
2330 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2333 void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
2335 /* lmv structures */
2337 __u32 ld_tgt_count; /* how many MDS's */
2338 __u32 ld_active_tgt_count; /* how many active */
2339 __u32 ld_default_stripe_count; /* how many objects are used */
2340 __u32 ld_pattern; /* default hash pattern */
2341 __u64 ld_default_hash_size;
2342 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2343 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2344 __u32 ld_qos_maxage; /* in second */
2345 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2346 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2347 struct obd_uuid ld_uuid;
2350 /* LMV layout EA, and it will be stored both in master and slave object */
2351 struct lmv_mds_md_v1 {
2353 __u32 lmv_stripe_count;
2354 __u32 lmv_master_mdt_index; /* On master object, it is master
2355 * MDT index, on slave object, it
2356 * is stripe index of the slave obj
2358 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2359 * which hash function to be used,
2360 * Note: only lower 16 bits is being
2361 * used for now. Higher 16 bits will
2362 * be used to mark the object status,
2363 * for example migrating or dead.
2365 __u32 lmv_layout_version; /* Used for directory restriping */
2369 char lmv_pool_name[LOV_MAXPOOLNAME + 1];/* pool name */
2370 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2373 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2374 #define LMV_MAGIC LMV_MAGIC_V1
2376 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2377 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2380 *Right now only the lower part(0-16bits) of lmv_hash_type is being used,
2381 * and the higher part will be the flag to indicate the status of object,
2382 * for example the object is being migrated. And the hash function
2383 * might be interpreted differently with different flags.
2385 #define LMV_HASH_TYPE_MASK 0x0000ffff
2387 #define LMV_HASH_FLAG_MIGRATION 0x80000000
2388 #define LMV_HASH_FLAG_DEAD 0x40000000
2391 * The FNV-1a hash algorithm is as follows:
2392 * hash = FNV_offset_basis
2393 * for each octet_of_data to be hashed
2394 * hash = hash XOR octet_of_data
2395 * hash = hash × FNV_prime
2397 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2399 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2400 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2402 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2403 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2404 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, size_t size)
2406 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2407 const unsigned char *p = buf;
2410 for (i = 0; i < size; i++) {
2412 hash *= LUSTRE_FNV_1A_64_PRIME;
2420 struct lmv_mds_md_v1 lmv_md_v1;
2421 struct lmv_user_md lmv_user_md;
2424 void lustre_swab_lmv_mds_md(union lmv_mds_md *lmm);
2426 static inline ssize_t lmv_mds_md_size(int stripe_count, unsigned int lmm_magic)
2428 ssize_t len = -EINVAL;
2430 switch (lmm_magic) {
2431 case LMV_MAGIC_V1: {
2432 struct lmv_mds_md_v1 *lmm1;
2434 len = sizeof(*lmm1);
2435 len += stripe_count * sizeof(lmm1->lmv_stripe_fids[0]);
2443 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2445 switch (le32_to_cpu(lmm->lmv_magic)) {
2447 return le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2448 case LMV_USER_MAGIC:
2449 return le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2455 static inline int lmv_mds_md_stripe_count_set(union lmv_mds_md *lmm,
2456 unsigned int stripe_count)
2460 switch (le32_to_cpu(lmm->lmv_magic)) {
2462 lmm->lmv_md_v1.lmv_stripe_count = cpu_to_le32(stripe_count);
2464 case LMV_USER_MAGIC:
2465 lmm->lmv_user_md.lum_stripe_count = cpu_to_le32(stripe_count);
2478 FLD_FIRST_OPC = FLD_QUERY
2484 SEQ_FIRST_OPC = SEQ_QUERY
2488 SEQ_ALLOC_SUPER = 0,
2499 * LOV data structures
2502 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2503 /* The size of the buffer the lov/mdc reserves for the
2504 * array of UUIDs returned by the MDS. With the current
2505 * protocol, this will limit the max number of OSTs per LOV
2508 #define LOV_DESC_MAGIC 0xB0CCDE5C
2509 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2510 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2512 /* LOV settings descriptor (should only contain static info) */
2514 __u32 ld_tgt_count; /* how many OBD's */
2515 __u32 ld_active_tgt_count; /* how many active */
2516 __u32 ld_default_stripe_count; /* how many objects are used */
2517 __u32 ld_pattern; /* default PATTERN_RAID0 */
2518 __u64 ld_default_stripe_size; /* in bytes */
2519 __u64 ld_default_stripe_offset; /* in bytes */
2520 __u32 ld_padding_0; /* unused */
2521 __u32 ld_qos_maxage; /* in second */
2522 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2523 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2524 struct obd_uuid ld_uuid;
2527 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2529 void lustre_swab_lov_desc(struct lov_desc *ld);
2534 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2539 LDLM_BL_CALLBACK = 104,
2540 LDLM_CP_CALLBACK = 105,
2541 LDLM_GL_CALLBACK = 106,
2542 LDLM_SET_INFO = 107,
2545 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2547 #define RES_NAME_SIZE 4
2548 struct ldlm_res_id {
2549 __u64 name[RES_NAME_SIZE];
2552 #define DLDLMRES "[%#llx:%#llx:%#llx].%llx"
2553 #define PLDLMRES(res) (res)->lr_name.name[0], (res)->lr_name.name[1], \
2554 (res)->lr_name.name[2], (res)->lr_name.name[3]
2556 static inline bool ldlm_res_eq(const struct ldlm_res_id *res0,
2557 const struct ldlm_res_id *res1)
2559 return !memcmp(res0, res1, sizeof(*res0));
2576 #define LCK_MODE_NUM 8
2586 #define LDLM_MIN_TYPE LDLM_PLAIN
2588 struct ldlm_extent {
2594 static inline int ldlm_extent_overlap(const struct ldlm_extent *ex1,
2595 const struct ldlm_extent *ex2)
2597 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
2600 /* check if @ex1 contains @ex2 */
2601 static inline int ldlm_extent_contain(const struct ldlm_extent *ex1,
2602 const struct ldlm_extent *ex2)
2604 return (ex1->start <= ex2->start) && (ex1->end >= ex2->end);
2607 struct ldlm_inodebits {
2611 struct ldlm_flock_wire {
2619 /* it's important that the fields of the ldlm_extent structure match
2620 * the first fields of the ldlm_flock structure because there is only
2621 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2622 * this ever changes we will need to swab the union differently based
2623 * on the resource type.
2626 union ldlm_wire_policy_data {
2627 struct ldlm_extent l_extent;
2628 struct ldlm_flock_wire l_flock;
2629 struct ldlm_inodebits l_inodebits;
2632 union ldlm_gl_desc {
2633 struct ldlm_gl_lquota_desc lquota_desc;
2636 void lustre_swab_gl_desc(union ldlm_gl_desc *);
2638 enum ldlm_intent_flags {
2641 IT_OPEN_CREAT = BIT(1) | BIT(0),
2642 IT_READDIR = BIT(2),
2643 IT_GETATTR = BIT(3),
2647 IT_GETXATTR = BIT(7),
2650 IT_LAYOUT = BIT(10),
2651 IT_QUOTA_DQACQ = BIT(11),
2652 IT_QUOTA_CONN = BIT(12),
2653 IT_SETXATTR = BIT(13),
2656 struct ldlm_intent {
2660 void lustre_swab_ldlm_intent(struct ldlm_intent *i);
2662 struct ldlm_resource_desc {
2663 enum ldlm_type lr_type;
2664 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
2665 struct ldlm_res_id lr_name;
2668 struct ldlm_lock_desc {
2669 struct ldlm_resource_desc l_resource;
2670 enum ldlm_mode l_req_mode;
2671 enum ldlm_mode l_granted_mode;
2672 union ldlm_wire_policy_data l_policy_data;
2675 #define LDLM_LOCKREQ_HANDLES 2
2676 #define LDLM_ENQUEUE_CANCEL_OFF 1
2678 struct ldlm_request {
2681 struct ldlm_lock_desc lock_desc;
2682 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2685 void lustre_swab_ldlm_request(struct ldlm_request *rq);
2687 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
2688 * Otherwise, 2 are available.
2690 #define ldlm_request_bufsize(count, type) \
2692 int _avail = LDLM_LOCKREQ_HANDLES; \
2693 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
2694 sizeof(struct ldlm_request) + \
2695 (count > _avail ? count - _avail : 0) * \
2696 sizeof(struct lustre_handle); \
2701 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2702 struct ldlm_lock_desc lock_desc;
2703 struct lustre_handle lock_handle;
2704 __u64 lock_policy_res1;
2705 __u64 lock_policy_res2;
2708 void lustre_swab_ldlm_reply(struct ldlm_reply *r);
2710 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2711 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2714 * Opcodes for mountconf (mgs and mgc)
2719 MGS_EXCEPTION, /* node died, etc. */
2720 MGS_TARGET_REG, /* whenever target starts up */
2726 #define MGS_FIRST_OPC MGS_CONNECT
2728 #define MGS_PARAM_MAXLEN 1024
2729 #define KEY_SET_INFO "set_info"
2731 struct mgs_send_param {
2732 char mgs_param[MGS_PARAM_MAXLEN];
2735 /* We pass this info to the MGS so it can write config logs */
2736 #define MTI_NAME_MAXLEN 64
2737 #define MTI_PARAM_MAXLEN 4096
2738 #define MTI_NIDS_MAX 32
2739 struct mgs_target_info {
2740 __u32 mti_lustre_ver;
2741 __u32 mti_stripe_index;
2742 __u32 mti_config_ver;
2744 __u32 mti_nid_count;
2745 __u32 mti_instance; /* Running instance of target */
2746 char mti_fsname[MTI_NAME_MAXLEN];
2747 char mti_svname[MTI_NAME_MAXLEN];
2748 char mti_uuid[sizeof(struct obd_uuid)];
2749 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
2750 char mti_params[MTI_PARAM_MAXLEN];
2753 void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
2755 struct mgs_nidtbl_entry {
2756 __u64 mne_version; /* table version of this entry */
2757 __u32 mne_instance; /* target instance # */
2758 __u32 mne_index; /* target index */
2759 __u32 mne_length; /* length of this entry - by bytes */
2760 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2761 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2762 __u8 mne_nid_size; /* size of each NID, by bytes */
2763 __u8 mne_nid_count; /* # of NIDs in buffer */
2765 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2769 void lustre_swab_mgs_nidtbl_entry(struct mgs_nidtbl_entry *oinfo);
2771 struct mgs_config_body {
2772 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2773 __u64 mcb_offset; /* next index of config log to request */
2774 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2776 __u8 mcb_bits; /* bits unit size of config log */
2777 __u32 mcb_units; /* # of units for bulk transfer */
2780 void lustre_swab_mgs_config_body(struct mgs_config_body *body);
2782 struct mgs_config_res {
2783 __u64 mcr_offset; /* index of last config log */
2784 __u64 mcr_size; /* size of the log */
2787 void lustre_swab_mgs_config_res(struct mgs_config_res *body);
2789 /* Config marker flags (in config log) */
2790 #define CM_START 0x01
2792 #define CM_SKIP 0x04
2793 #define CM_UPGRADE146 0x08
2794 #define CM_EXCLUDE 0x10
2795 #define CM_START_SKIP (CM_START | CM_SKIP)
2798 __u32 cm_step; /* aka config version */
2800 __u32 cm_vers; /* lustre release version number */
2801 __u32 cm_padding; /* 64 bit align */
2802 __s64 cm_createtime; /*when this record was first created */
2803 __s64 cm_canceltime; /*when this record is no longer valid*/
2804 char cm_tgtname[MTI_NAME_MAXLEN];
2805 char cm_comment[MTI_NAME_MAXLEN];
2808 void lustre_swab_cfg_marker(struct cfg_marker *marker, int swab, int size);
2811 * Opcodes for multiple servers.
2817 OBD_QC_CALLBACK, /* not used since 2.4 */
2821 #define OBD_FIRST_OPC OBD_PING
2824 * llog contexts indices.
2826 * There is compatibility problem with indexes below, they are not
2827 * continuous and must keep their numbers for compatibility needs.
2828 * See LU-5218 for details.
2831 LLOG_CONFIG_ORIG_CTXT = 0,
2832 LLOG_CONFIG_REPL_CTXT = 1,
2833 LLOG_MDS_OST_ORIG_CTXT = 2,
2834 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2835 LLOG_SIZE_ORIG_CTXT = 4,
2836 LLOG_SIZE_REPL_CTXT = 5,
2837 LLOG_TEST_ORIG_CTXT = 8,
2838 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2839 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2840 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2841 /* for multiple changelog consumers */
2842 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2843 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2847 /** Identifier for a single log object */
2849 struct ost_id lgl_oi;
2853 /** Records written to the CATALOGS list */
2854 #define CATLIST "CATALOGS"
2856 struct llog_logid lci_logid;
2862 /* Log data record types - there is no specific reason that these need to
2863 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2865 #define LLOG_OP_MAGIC 0x10600000
2866 #define LLOG_OP_MASK 0xfff00000
2869 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2870 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2871 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2872 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2873 REINT_UNLINK, /* obsolete after 2.5.0 */
2874 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2876 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2877 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2879 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2880 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2881 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2882 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2883 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2884 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2885 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2886 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2887 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2890 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2891 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2893 /** Log record header - stored in little endian order.
2894 * Each record must start with this struct, end with a llog_rec_tail,
2895 * and be a multiple of 256 bits in size.
2897 struct llog_rec_hdr {
2904 struct llog_rec_tail {
2909 /* Where data follow just after header */
2910 #define REC_DATA(ptr) \
2911 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2913 #define REC_DATA_LEN(rec) \
2914 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2915 sizeof(struct llog_rec_tail))
2917 struct llog_logid_rec {
2918 struct llog_rec_hdr lid_hdr;
2919 struct llog_logid lid_id;
2923 struct llog_rec_tail lid_tail;
2926 struct llog_unlink_rec {
2927 struct llog_rec_hdr lur_hdr;
2931 struct llog_rec_tail lur_tail;
2934 struct llog_unlink64_rec {
2935 struct llog_rec_hdr lur_hdr;
2936 struct lu_fid lur_fid;
2937 __u32 lur_count; /* to destroy the lost precreated */
2941 struct llog_rec_tail lur_tail;
2944 struct llog_setattr64_rec {
2945 struct llog_rec_hdr lsr_hdr;
2946 struct ost_id lsr_oi;
2952 struct llog_rec_tail lsr_tail;
2955 struct llog_size_change_rec {
2956 struct llog_rec_hdr lsc_hdr;
2957 struct ll_fid lsc_fid;
2962 struct llog_rec_tail lsc_tail;
2965 /* changelog llog name, needed by client replicators */
2966 #define CHANGELOG_CATALOG "changelog_catalog"
2968 struct changelog_setinfo {
2973 /** changelog record */
2974 struct llog_changelog_rec {
2975 struct llog_rec_hdr cr_hdr;
2976 struct changelog_rec cr; /**< Variable length field */
2977 struct llog_rec_tail cr_do_not_use; /**< for_sizezof_only */
2980 struct llog_changelog_user_rec {
2981 struct llog_rec_hdr cur_hdr;
2985 struct llog_rec_tail cur_tail;
2988 enum agent_req_status {
2996 static inline const char *agent_req_status2name(const enum agent_req_status ars)
3014 static inline bool agent_req_in_final_state(enum agent_req_status ars)
3016 return ((ars == ARS_SUCCEED) || (ars == ARS_FAILED) ||
3017 (ars == ARS_CANCELED));
3020 struct llog_agent_req_rec {
3021 struct llog_rec_hdr arr_hdr; /**< record header */
3022 __u32 arr_status; /**< status of the request */
3026 __u32 arr_archive_id; /**< backend archive number */
3027 __u64 arr_flags; /**< req flags */
3028 __u64 arr_compound_id;/**< compound cookie */
3029 __u64 arr_req_create; /**< req. creation time */
3030 __u64 arr_req_change; /**< req. status change time */
3031 struct hsm_action_item arr_hai; /**< req. to the agent */
3032 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
3035 /* Old llog gen for compatibility */
3041 struct llog_gen_rec {
3042 struct llog_rec_hdr lgr_hdr;
3043 struct llog_gen lgr_gen;
3047 struct llog_rec_tail lgr_tail;
3050 /* flags for the logs */
3052 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
3053 LLOG_F_IS_CAT = 0x2,
3054 LLOG_F_IS_PLAIN = 0x4,
3055 LLOG_F_EXT_JOBID = BIT(3),
3056 LLOG_F_IS_FIXSIZE = BIT(4),
3059 * Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
3060 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
3061 * because the catlog record is usually fixed size, but its plain
3062 * log record can be variable
3064 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID,
3067 /* On-disk header structure of each log object, stored in little endian order */
3068 #define LLOG_MIN_CHUNK_SIZE 8192
3069 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) +
3070 * sizeof(llh_tail) - sizeof(llh_bitmap)
3072 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
3073 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
3075 /* flags for the logs */
3076 struct llog_log_hdr {
3077 struct llog_rec_hdr llh_hdr;
3078 __s64 llh_timestamp;
3080 __u32 llh_bitmap_offset;
3084 /* for a catalog the first plain slot is next to it */
3085 struct obd_uuid llh_tgtuuid;
3086 __u32 llh_reserved[LLOG_HEADER_SIZE / sizeof(__u32) - 23];
3087 /* These fields must always be at the end of the llog_log_hdr.
3088 * Note: llh_bitmap size is variable because llog chunk size could be
3089 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
3090 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
3091 * llh_tail should only be referred by LLOG_HDR_TAIL().
3092 * But this structure is also used by client/server llog interface
3093 * (see llog_client.c), it will be kept in its original way to avoid
3094 * compatibility issue.
3096 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
3097 struct llog_rec_tail llh_tail;
3100 #undef LLOG_HEADER_SIZE
3101 #undef LLOG_BITMAP_BYTES
3103 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
3104 llh->llh_bitmap_offset - \
3105 sizeof(llh->llh_tail)) * 8)
3106 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
3107 (llh)->llh_bitmap_offset)
3108 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
3109 llh->llh_hdr.lrh_len - \
3110 sizeof(llh->llh_tail)))
3112 /** log cookies are used to reference a specific log file and a record
3115 struct llog_cookie {
3116 struct llog_logid lgc_lgl;
3122 /** llog protocol */
3123 enum llogd_rpc_ops {
3124 LLOG_ORIGIN_HANDLE_CREATE = 501,
3125 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
3126 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
3127 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
3128 LLOG_ORIGIN_HANDLE_CLOSE = 505,
3129 LLOG_ORIGIN_CONNECT = 506,
3130 LLOG_CATINFO = 507, /* deprecated */
3131 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
3132 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
3134 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
3138 struct llog_logid lgd_logid;
3140 __u32 lgd_llh_flags;
3142 __u32 lgd_saved_index;
3144 __u64 lgd_cur_offset;
3147 struct llogd_conn_body {
3148 struct llog_gen lgdc_gen;
3149 struct llog_logid lgdc_logid;
3150 __u32 lgdc_ctxt_idx;
3153 /* Note: 64-bit types are 64-bit aligned in structure */
3155 __u64 o_valid; /* hot fields in this obdo */
3158 __u64 o_size; /* o_size-o_blocks == ost_lvb */
3162 __u64 o_blocks; /* brw: cli sent cached bytes */
3165 /* 32-bit fields start here: keep an even number of them via padding */
3166 __u32 o_blksize; /* optimal IO blocksize */
3167 __u32 o_mode; /* brw: cli sent cache remain */
3171 __u32 o_nlink; /* brw: checksum */
3173 __u32 o_misc; /* brw: o_dropped */
3175 __u64 o_ioepoch; /* epoch in ost writes */
3176 __u32 o_stripe_idx; /* holds stripe idx */
3178 struct lustre_handle o_handle; /* brw: lock handle to prolong locks
3180 struct llog_cookie o_lcookie; /* destroy: unlink cookie from MDS,
3181 * obsolete in 2.8, reused in OSP
3186 __u64 o_data_version; /* getattr: sum of iversion for
3188 * brw: grant space consumed on
3189 * the client for the write
3196 #define o_dirty o_blocks
3197 #define o_undirty o_mode
3198 #define o_dropped o_misc
3199 #define o_cksum o_nlink
3200 #define o_grant_used o_data_version
3202 static inline void lustre_set_wire_obdo(const struct obd_connect_data *ocd,
3204 const struct obdo *lobdo)
3207 wobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3211 if (unlikely(!(ocd->ocd_connect_flags & OBD_CONNECT_FID)) &&
3212 fid_seq_is_echo(ostid_seq(&lobdo->o_oi))) {
3213 /* Currently OBD_FL_OSTID will only be used when 2.4 echo
3214 * client communicate with pre-2.4 server
3216 wobdo->o_oi.oi.oi_id = fid_oid(&lobdo->o_oi.oi_fid);
3217 wobdo->o_oi.oi.oi_seq = fid_seq(&lobdo->o_oi.oi_fid);
3221 static inline void lustre_get_wire_obdo(const struct obd_connect_data *ocd,
3223 const struct obdo *wobdo)
3225 __u32 local_flags = 0;
3227 if (lobdo->o_valid & OBD_MD_FLFLAGS)
3228 local_flags = lobdo->o_flags & OBD_FL_LOCAL_MASK;
3231 if (local_flags != 0) {
3232 lobdo->o_valid |= OBD_MD_FLFLAGS;
3233 lobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3234 lobdo->o_flags |= local_flags;
3239 if (unlikely(!(ocd->ocd_connect_flags & OBD_CONNECT_FID)) &&
3240 fid_seq_is_echo(wobdo->o_oi.oi.oi_seq)) {
3242 lobdo->o_oi.oi_fid.f_seq = wobdo->o_oi.oi.oi_seq;
3243 lobdo->o_oi.oi_fid.f_oid = wobdo->o_oi.oi.oi_id;
3244 lobdo->o_oi.oi_fid.f_ver = 0;
3248 /* request structure for OST's */
3253 /* Key for FIEMAP to be used in get_info calls */
3254 struct ll_fiemap_info_key {
3256 struct obdo lfik_oa;
3257 struct fiemap lfik_fiemap;
3260 void lustre_swab_ost_body(struct ost_body *b);
3261 void lustre_swab_ost_last_id(__u64 *id);
3262 void lustre_swab_fiemap(struct fiemap *fiemap);
3264 void lustre_swab_lov_user_md_v1(struct lov_user_md_v1 *lum);
3265 void lustre_swab_lov_user_md_v3(struct lov_user_md_v3 *lum);
3266 void lustre_swab_lov_user_md_objects(struct lov_user_ost_data *lod,
3268 void lustre_swab_lov_mds_md(struct lov_mds_md *lmm);
3271 void lustre_swab_llogd_body(struct llogd_body *d);
3272 void lustre_swab_llog_hdr(struct llog_log_hdr *h);
3273 void lustre_swab_llogd_conn_body(struct llogd_conn_body *d);
3274 void lustre_swab_llog_rec(struct llog_rec_hdr *rec);
3277 void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
3279 /* Functions for dumping PTLRPC fields */
3280 void dump_rniobuf(struct niobuf_remote *rnb);
3281 void dump_ioo(struct obd_ioobj *nb);
3282 void dump_ost_body(struct ost_body *ob);
3283 void dump_rcs(__u32 *rc);
3285 /* security opcodes */
3288 SEC_CTX_INIT_CONT = 802,
3291 SEC_FIRST_OPC = SEC_CTX_INIT
3295 * capa related definitions
3297 #define CAPA_HMAC_MAX_LEN 64
3298 #define CAPA_HMAC_KEY_MAX_LEN 56
3300 /* NB take care when changing the sequence of elements this struct,
3301 * because the offset info is used in find_capa()
3303 struct lustre_capa {
3304 struct lu_fid lc_fid; /** fid */
3305 __u64 lc_opc; /** operations allowed */
3306 __u64 lc_uid; /** file owner */
3307 __u64 lc_gid; /** file group */
3308 __u32 lc_flags; /** HMAC algorithm & flags */
3309 __u32 lc_keyid; /** key# used for the capability */
3310 __u32 lc_timeout; /** capa timeout value (sec) */
3311 /* FIXME: y2038 time_t overflow: */
3312 __u32 lc_expiry; /** expiry time (sec) */
3313 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
3316 void lustre_swab_lustre_capa(struct lustre_capa *c);
3318 /** lustre_capa::lc_opc */
3320 CAPA_OPC_BODY_WRITE = 1 << 0, /**< write object data */
3321 CAPA_OPC_BODY_READ = 1 << 1, /**< read object data */
3322 CAPA_OPC_INDEX_LOOKUP = 1 << 2, /**< lookup object fid */
3323 CAPA_OPC_INDEX_INSERT = 1 << 3, /**< insert object fid */
3324 CAPA_OPC_INDEX_DELETE = 1 << 4, /**< delete object fid */
3325 CAPA_OPC_OSS_WRITE = 1 << 5, /**< write oss object data */
3326 CAPA_OPC_OSS_READ = 1 << 6, /**< read oss object data */
3327 CAPA_OPC_OSS_TRUNC = 1 << 7, /**< truncate oss object */
3328 CAPA_OPC_OSS_DESTROY = 1 << 8, /**< destroy oss object */
3329 CAPA_OPC_META_WRITE = 1 << 9, /**< write object meta data */
3330 CAPA_OPC_META_READ = 1 << 10, /**< read object meta data */
3333 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3334 #define CAPA_OPC_MDS_ONLY \
3335 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3336 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3337 #define CAPA_OPC_OSS_ONLY \
3338 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3339 CAPA_OPC_OSS_DESTROY)
3340 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3341 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3343 struct lustre_capa_key {
3344 __u64 lk_seq; /**< mds# */
3345 __u32 lk_keyid; /**< key# */
3347 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
3350 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3351 #define LINK_EA_MAGIC 0x11EAF1DFUL
3352 struct link_ea_header {
3355 __u64 leh_len; /* total size */
3361 /** Hardlink data is name and parent fid.
3362 * Stored in this crazy struct for maximum packing and endian-neutrality
3364 struct link_ea_entry {
3365 /** __u16 stored big-endian, unaligned */
3366 unsigned char lee_reclen[2];
3367 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3371 /** fid2path request/reply structure */
3372 struct getinfo_fid2path {
3373 struct lu_fid gf_fid;
3380 void lustre_swab_fid2path(struct getinfo_fid2path *gf);
3382 /** path2parent request/reply structures */
3384 struct lu_fid gp_fid; /**< parent FID */
3385 __u32 gp_linkno; /**< hardlink number */
3386 __u32 gp_name_size; /**< size of the name field */
3387 char gp_name[0]; /**< zero-terminated link name */
3391 LAYOUT_INTENT_ACCESS = 0,
3392 LAYOUT_INTENT_READ = 1,
3393 LAYOUT_INTENT_WRITE = 2,
3394 LAYOUT_INTENT_GLIMPSE = 3,
3395 LAYOUT_INTENT_TRUNC = 4,
3396 LAYOUT_INTENT_RELEASE = 5,
3397 LAYOUT_INTENT_RESTORE = 6
3400 /* enqueue layout lock with intent */
3401 struct layout_intent {
3402 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3408 void lustre_swab_layout_intent(struct layout_intent *li);
3411 * On the wire version of hsm_progress structure.
3413 * Contains the userspace hsm_progress and some internal fields.
3415 struct hsm_progress_kernel {
3416 /* Field taken from struct hsm_progress */
3417 struct lu_fid hpk_fid;
3419 struct hsm_extent hpk_extent;
3421 __u16 hpk_errval; /* positive val */
3423 /* Additional fields */
3424 __u64 hpk_data_version;
3428 void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
3429 void lustre_swab_hsm_current_action(struct hsm_current_action *action);
3430 void lustre_swab_hsm_progress_kernel(struct hsm_progress_kernel *hpk);
3431 void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
3432 void lustre_swab_hsm_user_item(struct hsm_user_item *hui);
3433 void lustre_swab_hsm_request(struct hsm_request *hr);
3435 /** layout swap request structure
3436 * fid1 and fid2 are in mdt_body
3438 struct mdc_swap_layouts {
3442 void lustre_swab_swap_layouts(struct mdc_swap_layouts *msl);
3445 struct lustre_handle cd_handle;
3446 struct lu_fid cd_fid;
3447 __u64 cd_data_version;
3448 __u64 cd_reserved[8];
3451 void lustre_swab_close_data(struct close_data *data);