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"
100 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
101 * FOO_REPLY_PORTAL is for incoming replies on the FOO
102 * FOO_BULK_PORTAL is for incoming bulk on the FOO
105 /* Lustre service names are following the format
106 * service name + MDT + seq name
108 #define LUSTRE_MDT_MAXNAMELEN 80
110 #define CONNMGR_REQUEST_PORTAL 1
111 #define CONNMGR_REPLY_PORTAL 2
112 /*#define OSC_REQUEST_PORTAL 3 */
113 #define OSC_REPLY_PORTAL 4
114 /*#define OSC_BULK_PORTAL 5 */
115 #define OST_IO_PORTAL 6
116 #define OST_CREATE_PORTAL 7
117 #define OST_BULK_PORTAL 8
118 /*#define MDC_REQUEST_PORTAL 9 */
119 #define MDC_REPLY_PORTAL 10
120 /*#define MDC_BULK_PORTAL 11 */
121 #define MDS_REQUEST_PORTAL 12
122 /*#define MDS_REPLY_PORTAL 13 */
123 #define MDS_BULK_PORTAL 14
124 #define LDLM_CB_REQUEST_PORTAL 15
125 #define LDLM_CB_REPLY_PORTAL 16
126 #define LDLM_CANCEL_REQUEST_PORTAL 17
127 #define LDLM_CANCEL_REPLY_PORTAL 18
128 /*#define PTLBD_REQUEST_PORTAL 19 */
129 /*#define PTLBD_REPLY_PORTAL 20 */
130 /*#define PTLBD_BULK_PORTAL 21 */
131 #define MDS_SETATTR_PORTAL 22
132 #define MDS_READPAGE_PORTAL 23
133 #define OUT_PORTAL 24
135 #define MGC_REPLY_PORTAL 25
136 #define MGS_REQUEST_PORTAL 26
137 #define MGS_REPLY_PORTAL 27
138 #define OST_REQUEST_PORTAL 28
139 #define FLD_REQUEST_PORTAL 29
140 #define SEQ_METADATA_PORTAL 30
141 #define SEQ_DATA_PORTAL 31
142 #define SEQ_CONTROLLER_PORTAL 32
143 #define MGS_BULK_PORTAL 33
145 /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com,
150 #define PTL_RPC_MSG_REQUEST 4711
151 #define PTL_RPC_MSG_ERR 4712
152 #define PTL_RPC_MSG_REPLY 4713
154 /* DON'T use swabbed values of MAGIC as magic! */
155 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
156 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
158 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
160 #define PTLRPC_MSG_VERSION 0x00000003
161 #define LUSTRE_VERSION_MASK 0xffff0000
162 #define LUSTRE_OBD_VERSION 0x00010000
163 #define LUSTRE_MDS_VERSION 0x00020000
164 #define LUSTRE_OST_VERSION 0x00030000
165 #define LUSTRE_DLM_VERSION 0x00040000
166 #define LUSTRE_LOG_VERSION 0x00050000
167 #define LUSTRE_MGS_VERSION 0x00060000
170 * Describes a range of sequence, lsr_start is included but lsr_end is
172 * Same structure is used in fld module where lsr_index field holds mdt id
175 struct lu_seq_range
{
182 struct lu_seq_range_array
{
185 struct lu_seq_range lsra_lsr
[0];
188 #define LU_SEQ_RANGE_MDT 0x0
189 #define LU_SEQ_RANGE_OST 0x1
190 #define LU_SEQ_RANGE_ANY 0x3
192 #define LU_SEQ_RANGE_MASK 0x3
194 static inline unsigned fld_range_type(const struct lu_seq_range
*range
)
196 return range
->lsr_flags
& LU_SEQ_RANGE_MASK
;
199 static inline bool fld_range_is_ost(const struct lu_seq_range
*range
)
201 return fld_range_type(range
) == LU_SEQ_RANGE_OST
;
204 static inline bool fld_range_is_mdt(const struct lu_seq_range
*range
)
206 return fld_range_type(range
) == LU_SEQ_RANGE_MDT
;
210 * This all range is only being used when fld client sends fld query request,
211 * but it does not know whether the seq is MDT or OST, so it will send req
212 * with ALL type, which means either seq type gotten from lookup can be
215 static inline unsigned fld_range_is_any(const struct lu_seq_range
*range
)
217 return fld_range_type(range
) == LU_SEQ_RANGE_ANY
;
220 static inline void fld_range_set_type(struct lu_seq_range
*range
,
223 range
->lsr_flags
|= flags
;
226 static inline void fld_range_set_mdt(struct lu_seq_range
*range
)
228 fld_range_set_type(range
, LU_SEQ_RANGE_MDT
);
231 static inline void fld_range_set_ost(struct lu_seq_range
*range
)
233 fld_range_set_type(range
, LU_SEQ_RANGE_OST
);
236 static inline void fld_range_set_any(struct lu_seq_range
*range
)
238 fld_range_set_type(range
, LU_SEQ_RANGE_ANY
);
242 * returns width of given range \a r
245 static inline __u64
range_space(const struct lu_seq_range
*range
)
247 return range
->lsr_end
- range
->lsr_start
;
251 * initialize range to zero
254 static inline void range_init(struct lu_seq_range
*range
)
256 memset(range
, 0, sizeof(*range
));
260 * check if given seq id \a s is within given range \a r
263 static inline bool range_within(const struct lu_seq_range
*range
,
266 return s
>= range
->lsr_start
&& s
< range
->lsr_end
;
269 static inline bool range_is_sane(const struct lu_seq_range
*range
)
271 return (range
->lsr_end
>= range
->lsr_start
);
274 static inline bool range_is_zero(const struct lu_seq_range
*range
)
276 return (range
->lsr_start
== 0 && range
->lsr_end
== 0);
279 static inline bool range_is_exhausted(const struct lu_seq_range
*range
)
282 return range_space(range
) == 0;
285 /* return 0 if two range have the same location */
286 static inline int range_compare_loc(const struct lu_seq_range
*r1
,
287 const struct lu_seq_range
*r2
)
289 return r1
->lsr_index
!= r2
->lsr_index
||
290 r1
->lsr_flags
!= r2
->lsr_flags
;
293 #define DRANGE "[%#16.16Lx-%#16.16Lx):%x:%s"
295 #define PRANGE(range) \
296 (range)->lsr_start, \
298 (range)->lsr_index, \
299 fld_range_is_mdt(range) ? "mdt" : "ost"
301 /** \defgroup lu_fid lu_fid
306 * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat.
307 * Deprecated since HSM and SOM attributes are now stored in separate on-disk
311 LMAC_HSM
= 0x00000001,
312 LMAC_SOM
= 0x00000002,
313 LMAC_NOT_IN_OI
= 0x00000004, /* the object does NOT need OI mapping */
314 LMAC_FID_ON_OST
= 0x00000008, /* For OST-object, its OI mapping is
315 * under /O/<seq>/d<x>.
320 * Masks for all features that should be supported by a Lustre version to
321 * access a specific file.
322 * This information is stored in lustre_mdt_attrs::lma_incompat.
325 LMAI_RELEASED
= 0x00000001, /* file is released */
326 LMAI_AGENT
= 0x00000002, /* agent inode */
327 LMAI_REMOTE_PARENT
= 0x00000004, /* the parent of the object
328 * is on the remote MDT
332 #define LMA_INCOMPAT_SUPP (LMAI_AGENT | LMAI_REMOTE_PARENT)
338 /** LASTID file has zero OID */
339 LUSTRE_FID_LASTID_OID
= 0UL,
340 /** initial fid id value */
341 LUSTRE_FID_INIT_OID
= 1UL
344 /** returns fid object sequence */
345 static inline __u64
fid_seq(const struct lu_fid
*fid
)
350 /** returns fid object id */
351 static inline __u32
fid_oid(const struct lu_fid
*fid
)
356 /** returns fid object version */
357 static inline __u32
fid_ver(const struct lu_fid
*fid
)
362 static inline void fid_zero(struct lu_fid
*fid
)
364 memset(fid
, 0, sizeof(*fid
));
367 static inline __u64
fid_ver_oid(const struct lu_fid
*fid
)
369 return ((__u64
)fid_ver(fid
) << 32 | fid_oid(fid
));
372 /* copytool uses a 32b bitmask field to encode archive-Ids during register
374 * archive num = 0 => all
375 * archive num from 1 to 32
377 #define LL_HSM_MAX_ARCHIVE (sizeof(__u32) * 8)
380 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
381 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
382 * used for other purposes and not risk collisions with existing inodes.
384 * Different FID Format
385 * http://wiki.old.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs
388 FID_SEQ_OST_MDT0
= 0,
389 FID_SEQ_LLOG
= 1, /* unnamed llogs */
391 FID_SEQ_OST_MDT1
= 3,
392 FID_SEQ_OST_MAX
= 9, /* Max MDT count before OST_on_FID */
393 FID_SEQ_LLOG_NAME
= 10, /* named llogs */
396 FID_SEQ_IGIF_MAX
= 0x0ffffffffULL
,
397 FID_SEQ_IDIF
= 0x100000000ULL
,
398 FID_SEQ_IDIF_MAX
= 0x1ffffffffULL
,
399 /* Normal FID sequence starts from this value, i.e. 1<<33 */
400 FID_SEQ_START
= 0x200000000ULL
,
401 /* sequence for local pre-defined FIDs listed in local_oid */
402 FID_SEQ_LOCAL_FILE
= 0x200000001ULL
,
403 FID_SEQ_DOT_LUSTRE
= 0x200000002ULL
,
404 /* sequence is used for local named objects FIDs generated
405 * by local_object_storage library
407 FID_SEQ_LOCAL_NAME
= 0x200000003ULL
,
408 /* Because current FLD will only cache the fid sequence, instead
409 * of oid on the client side, if the FID needs to be exposed to
410 * clients sides, it needs to make sure all of fids under one
411 * sequence will be located in one MDT.
413 FID_SEQ_SPECIAL
= 0x200000004ULL
,
414 FID_SEQ_QUOTA
= 0x200000005ULL
,
415 FID_SEQ_QUOTA_GLB
= 0x200000006ULL
,
416 FID_SEQ_ROOT
= 0x200000007ULL
, /* Located on MDT0 */
417 FID_SEQ_NORMAL
= 0x200000400ULL
,
418 FID_SEQ_LOV_DEFAULT
= 0xffffffffffffffffULL
421 #define OBIF_OID_MAX_BITS 32
422 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
423 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
424 #define IDIF_OID_MAX_BITS 48
425 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
426 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
428 /** OID for FID_SEQ_SPECIAL */
430 /* Big Filesystem Lock to serialize rename operations */
431 FID_OID_SPECIAL_BFL
= 1UL,
434 /** OID for FID_SEQ_DOT_LUSTRE */
435 enum dot_lustre_oid
{
436 FID_OID_DOT_LUSTRE
= 1UL,
437 FID_OID_DOT_LUSTRE_OBF
= 2UL,
440 static inline bool fid_seq_is_mdt0(__u64 seq
)
442 return (seq
== FID_SEQ_OST_MDT0
);
445 static inline bool fid_seq_is_mdt(__u64 seq
)
447 return seq
== FID_SEQ_OST_MDT0
|| seq
>= FID_SEQ_NORMAL
;
450 static inline bool fid_seq_is_echo(__u64 seq
)
452 return (seq
== FID_SEQ_ECHO
);
455 static inline bool fid_is_echo(const struct lu_fid
*fid
)
457 return fid_seq_is_echo(fid_seq(fid
));
460 static inline bool fid_seq_is_llog(__u64 seq
)
462 return (seq
== FID_SEQ_LLOG
);
465 static inline bool fid_is_llog(const struct lu_fid
*fid
)
467 /* file with OID == 0 is not llog but contains last oid */
468 return fid_seq_is_llog(fid_seq(fid
)) && fid_oid(fid
) > 0;
471 static inline bool fid_seq_is_rsvd(__u64 seq
)
473 return (seq
> FID_SEQ_OST_MDT0
&& seq
<= FID_SEQ_RSVD
);
476 static inline bool fid_seq_is_special(__u64 seq
)
478 return seq
== FID_SEQ_SPECIAL
;
481 static inline bool fid_seq_is_local_file(__u64 seq
)
483 return seq
== FID_SEQ_LOCAL_FILE
||
484 seq
== FID_SEQ_LOCAL_NAME
;
487 static inline bool fid_seq_is_root(__u64 seq
)
489 return seq
== FID_SEQ_ROOT
;
492 static inline bool fid_seq_is_dot(__u64 seq
)
494 return seq
== FID_SEQ_DOT_LUSTRE
;
497 static inline bool fid_seq_is_default(__u64 seq
)
499 return seq
== FID_SEQ_LOV_DEFAULT
;
502 static inline bool fid_is_mdt0(const struct lu_fid
*fid
)
504 return fid_seq_is_mdt0(fid_seq(fid
));
507 static inline void lu_root_fid(struct lu_fid
*fid
)
509 fid
->f_seq
= FID_SEQ_ROOT
;
515 * Check if a fid is igif or not.
516 * \param fid the fid to be tested.
517 * \return true if the fid is a igif; otherwise false.
519 static inline bool fid_seq_is_igif(__u64 seq
)
521 return seq
>= FID_SEQ_IGIF
&& seq
<= FID_SEQ_IGIF_MAX
;
524 static inline bool fid_is_igif(const struct lu_fid
*fid
)
526 return fid_seq_is_igif(fid_seq(fid
));
530 * Check if a fid is idif or not.
531 * \param fid the fid to be tested.
532 * \return true if the fid is a idif; otherwise false.
534 static inline bool fid_seq_is_idif(__u64 seq
)
536 return seq
>= FID_SEQ_IDIF
&& seq
<= FID_SEQ_IDIF_MAX
;
539 static inline bool fid_is_idif(const struct lu_fid
*fid
)
541 return fid_seq_is_idif(fid_seq(fid
));
544 static inline bool fid_is_local_file(const struct lu_fid
*fid
)
546 return fid_seq_is_local_file(fid_seq(fid
));
549 static inline bool fid_seq_is_norm(__u64 seq
)
551 return (seq
>= FID_SEQ_NORMAL
);
554 static inline bool fid_is_norm(const struct lu_fid
*fid
)
556 return fid_seq_is_norm(fid_seq(fid
));
559 /* convert an OST objid into an IDIF FID SEQ number */
560 static inline __u64
fid_idif_seq(__u64 id
, __u32 ost_idx
)
562 return FID_SEQ_IDIF
| (ost_idx
<< 16) | ((id
>> 32) & 0xffff);
565 /* convert a packed IDIF FID into an OST objid */
566 static inline __u64
fid_idif_id(__u64 seq
, __u32 oid
, __u32 ver
)
568 return ((__u64
)ver
<< 48) | ((seq
& 0xffff) << 32) | oid
;
571 /* extract ost index from IDIF FID */
572 static inline __u32
fid_idif_ost_idx(const struct lu_fid
*fid
)
574 return (fid_seq(fid
) >> 16) & 0xffff;
577 /* extract OST sequence (group) from a wire ost_id (id/seq) pair */
578 static inline __u64
ostid_seq(const struct ost_id
*ostid
)
580 if (fid_seq_is_mdt0(ostid
->oi
.oi_seq
))
581 return FID_SEQ_OST_MDT0
;
583 if (unlikely(fid_seq_is_default(ostid
->oi
.oi_seq
)))
584 return FID_SEQ_LOV_DEFAULT
;
586 if (fid_is_idif(&ostid
->oi_fid
))
587 return FID_SEQ_OST_MDT0
;
589 return fid_seq(&ostid
->oi_fid
);
592 /* extract OST objid from a wire ost_id (id/seq) pair */
593 static inline __u64
ostid_id(const struct ost_id
*ostid
)
595 if (fid_seq_is_mdt0(ostid
->oi
.oi_seq
))
596 return ostid
->oi
.oi_id
& IDIF_OID_MASK
;
598 if (unlikely(fid_seq_is_default(ostid
->oi
.oi_seq
)))
599 return ostid
->oi
.oi_id
;
601 if (fid_is_idif(&ostid
->oi_fid
))
602 return fid_idif_id(fid_seq(&ostid
->oi_fid
),
603 fid_oid(&ostid
->oi_fid
), 0);
605 return fid_oid(&ostid
->oi_fid
);
608 static inline void ostid_set_seq(struct ost_id
*oi
, __u64 seq
)
610 if (fid_seq_is_mdt0(seq
) || fid_seq_is_default(seq
)) {
613 oi
->oi_fid
.f_seq
= seq
;
614 /* Note: if f_oid + f_ver is zero, we need init it
615 * to be 1, otherwise, ostid_seq will treat this
616 * as old ostid (oi_seq == 0)
618 if (oi
->oi_fid
.f_oid
== 0 && oi
->oi_fid
.f_ver
== 0)
619 oi
->oi_fid
.f_oid
= LUSTRE_FID_INIT_OID
;
623 static inline void ostid_set_seq_mdt0(struct ost_id
*oi
)
625 ostid_set_seq(oi
, FID_SEQ_OST_MDT0
);
628 static inline void ostid_set_seq_echo(struct ost_id
*oi
)
630 ostid_set_seq(oi
, FID_SEQ_ECHO
);
633 static inline void ostid_set_seq_llog(struct ost_id
*oi
)
635 ostid_set_seq(oi
, FID_SEQ_LLOG
);
639 * Note: we need check oi_seq to decide where to set oi_id,
640 * so oi_seq should always be set ahead of oi_id.
642 static inline void ostid_set_id(struct ost_id
*oi
, __u64 oid
)
644 if (fid_seq_is_mdt0(oi
->oi
.oi_seq
)) {
645 if (oid
>= IDIF_MAX_OID
) {
646 CERROR("Bad %llu to set " DOSTID
"\n", oid
, POSTID(oi
));
650 } else if (fid_is_idif(&oi
->oi_fid
)) {
651 if (oid
>= IDIF_MAX_OID
) {
652 CERROR("Bad %llu to set "DOSTID
"\n",
656 oi
->oi_fid
.f_seq
= fid_idif_seq(oid
,
657 fid_idif_ost_idx(&oi
->oi_fid
));
658 oi
->oi_fid
.f_oid
= oid
;
659 oi
->oi_fid
.f_ver
= oid
>> 48;
661 if (oid
> OBIF_MAX_OID
) {
662 CERROR("Bad %llu to set " DOSTID
"\n", oid
, POSTID(oi
));
665 oi
->oi_fid
.f_oid
= oid
;
669 static inline int fid_set_id(struct lu_fid
*fid
, __u64 oid
)
671 if (unlikely(fid_seq_is_igif(fid
->f_seq
))) {
672 CERROR("bad IGIF, "DFID
"\n", PFID(fid
));
676 if (fid_is_idif(fid
)) {
677 if (oid
>= IDIF_MAX_OID
) {
678 CERROR("Too large OID %#llx to set IDIF "DFID
"\n",
679 (unsigned long long)oid
, PFID(fid
));
682 fid
->f_seq
= fid_idif_seq(oid
, fid_idif_ost_idx(fid
));
684 fid
->f_ver
= oid
>> 48;
686 if (oid
> OBIF_MAX_OID
) {
687 CERROR("Too large OID %#llx to set REG "DFID
"\n",
688 (unsigned long long)oid
, PFID(fid
));
697 * Unpack an OST object id/seq (group) into a FID. This is needed for
698 * converting all obdo, lmm, lsm, etc. 64-bit id/seq pairs into proper
699 * FIDs. Note that if an id/seq is already in FID/IDIF format it will
700 * be passed through unchanged. Only legacy OST objects in "group 0"
701 * will be mapped into the IDIF namespace so that they can fit into the
702 * struct lu_fid fields without loss. For reference see:
703 * http://wiki.old.lustre.org/index.php/Architecture_-_Interoperability_fids_zfs
705 static inline int ostid_to_fid(struct lu_fid
*fid
, struct ost_id
*ostid
,
708 __u64 seq
= ostid_seq(ostid
);
710 if (ost_idx
> 0xffff) {
711 CERROR("bad ost_idx, "DOSTID
" ost_idx:%u\n", POSTID(ostid
),
716 if (fid_seq_is_mdt0(seq
)) {
717 __u64 oid
= ostid_id(ostid
);
719 /* This is a "legacy" (old 1.x/2.early) OST object in "group 0"
720 * that we map into the IDIF namespace. It allows up to 2^48
721 * objects per OST, as this is the object namespace that has
722 * been in production for years. This can handle create rates
723 * of 1M objects/s/OST for 9 years, or combinations thereof.
725 if (oid
>= IDIF_MAX_OID
) {
726 CERROR("bad MDT0 id, " DOSTID
" ost_idx:%u\n",
727 POSTID(ostid
), ost_idx
);
730 fid
->f_seq
= fid_idif_seq(oid
, ost_idx
);
731 /* truncate to 32 bits by assignment */
733 /* in theory, not currently used */
734 fid
->f_ver
= oid
>> 48;
735 } else if (likely(!fid_seq_is_default(seq
))) {
736 /* This is either an IDIF object, which identifies objects across
737 * all OSTs, or a regular FID. The IDIF namespace maps legacy
738 * OST objects into the FID namespace. In both cases, we just
739 * pass the FID through, no conversion needed.
741 if (ostid
->oi_fid
.f_ver
!= 0) {
742 CERROR("bad MDT0 id, " DOSTID
" ost_idx:%u\n",
743 POSTID(ostid
), ost_idx
);
746 *fid
= ostid
->oi_fid
;
752 /* pack any OST FID into an ostid (id/seq) for the wire/disk */
753 static inline int fid_to_ostid(const struct lu_fid
*fid
, struct ost_id
*ostid
)
755 if (unlikely(fid_seq_is_igif(fid
->f_seq
))) {
756 CERROR("bad IGIF, "DFID
"\n", PFID(fid
));
760 if (fid_is_idif(fid
)) {
761 ostid_set_seq_mdt0(ostid
);
762 ostid_set_id(ostid
, fid_idif_id(fid_seq(fid
), fid_oid(fid
),
765 ostid
->oi_fid
= *fid
;
771 /* Check whether the fid is for LAST_ID */
772 static inline bool fid_is_last_id(const struct lu_fid
*fid
)
774 return (fid_oid(fid
) == 0);
778 * Get inode number from a igif.
779 * \param fid a igif to get inode number from.
780 * \return inode number for the igif.
782 static inline ino_t
lu_igif_ino(const struct lu_fid
*fid
)
787 void lustre_swab_ost_id(struct ost_id
*oid
);
790 * Get inode generation from a igif.
791 * \param fid a igif to get inode generation from.
792 * \return inode generation for the igif.
794 static inline __u32
lu_igif_gen(const struct lu_fid
*fid
)
800 * Build igif from the inode number/generation.
802 static inline void lu_igif_build(struct lu_fid
*fid
, __u32 ino
, __u32 gen
)
810 * Fids are transmitted across network (in the sender byte-ordering),
811 * and stored on disk in big-endian order.
813 static inline void fid_cpu_to_le(struct lu_fid
*dst
, const struct lu_fid
*src
)
815 dst
->f_seq
= cpu_to_le64(fid_seq(src
));
816 dst
->f_oid
= cpu_to_le32(fid_oid(src
));
817 dst
->f_ver
= cpu_to_le32(fid_ver(src
));
820 static inline void fid_le_to_cpu(struct lu_fid
*dst
, const struct lu_fid
*src
)
822 dst
->f_seq
= le64_to_cpu(fid_seq(src
));
823 dst
->f_oid
= le32_to_cpu(fid_oid(src
));
824 dst
->f_ver
= le32_to_cpu(fid_ver(src
));
827 static inline void fid_cpu_to_be(struct lu_fid
*dst
, const struct lu_fid
*src
)
829 dst
->f_seq
= cpu_to_be64(fid_seq(src
));
830 dst
->f_oid
= cpu_to_be32(fid_oid(src
));
831 dst
->f_ver
= cpu_to_be32(fid_ver(src
));
834 static inline void fid_be_to_cpu(struct lu_fid
*dst
, const struct lu_fid
*src
)
836 dst
->f_seq
= be64_to_cpu(fid_seq(src
));
837 dst
->f_oid
= be32_to_cpu(fid_oid(src
));
838 dst
->f_ver
= be32_to_cpu(fid_ver(src
));
841 static inline bool fid_is_sane(const struct lu_fid
*fid
)
844 ((fid_seq(fid
) >= FID_SEQ_START
&& fid_ver(fid
) == 0) ||
845 fid_is_igif(fid
) || fid_is_idif(fid
) ||
846 fid_seq_is_rsvd(fid_seq(fid
)));
849 static inline bool fid_is_zero(const struct lu_fid
*fid
)
851 return fid_seq(fid
) == 0 && fid_oid(fid
) == 0;
854 void lustre_swab_lu_fid(struct lu_fid
*fid
);
855 void lustre_swab_lu_seq_range(struct lu_seq_range
*range
);
857 static inline bool lu_fid_eq(const struct lu_fid
*f0
, const struct lu_fid
*f1
)
859 return memcmp(f0
, f1
, sizeof(*f0
)) == 0;
862 #define __diff_normalize(val0, val1) \
864 typeof(val0) __val0 = (val0); \
865 typeof(val1) __val1 = (val1); \
867 (__val0 == __val1 ? 0 : __val0 > __val1 ? 1 : -1); \
870 static inline int lu_fid_cmp(const struct lu_fid
*f0
,
871 const struct lu_fid
*f1
)
874 __diff_normalize(fid_seq(f0
), fid_seq(f1
)) ?:
875 __diff_normalize(fid_oid(f0
), fid_oid(f1
)) ?:
876 __diff_normalize(fid_ver(f0
), fid_ver(f1
));
879 static inline void ostid_cpu_to_le(const struct ost_id
*src_oi
,
880 struct ost_id
*dst_oi
)
882 if (fid_seq_is_mdt0(ostid_seq(src_oi
))) {
883 dst_oi
->oi
.oi_id
= cpu_to_le64(src_oi
->oi
.oi_id
);
884 dst_oi
->oi
.oi_seq
= cpu_to_le64(src_oi
->oi
.oi_seq
);
886 fid_cpu_to_le(&dst_oi
->oi_fid
, &src_oi
->oi_fid
);
890 static inline void ostid_le_to_cpu(const struct ost_id
*src_oi
,
891 struct ost_id
*dst_oi
)
893 if (fid_seq_is_mdt0(ostid_seq(src_oi
))) {
894 dst_oi
->oi
.oi_id
= le64_to_cpu(src_oi
->oi
.oi_id
);
895 dst_oi
->oi
.oi_seq
= le64_to_cpu(src_oi
->oi
.oi_seq
);
897 fid_le_to_cpu(&dst_oi
->oi_fid
, &src_oi
->oi_fid
);
903 /** \defgroup lu_dir lu_dir
908 * Enumeration of possible directory entry attributes.
910 * Attributes follow directory entry header in the order they appear in this
913 enum lu_dirent_attrs
{
916 LUDA_64BITHASH
= 0x0004,
920 * Layout of readdir pages, as transmitted on wire.
923 /** valid if LUDA_FID is set. */
924 struct lu_fid lde_fid
;
925 /** a unique entry identifier: a hash or an offset. */
927 /** total record length, including all attributes. */
931 /** optional variable size attributes following this entry.
932 * taken from enum lu_dirent_attrs.
935 /** name is followed by the attributes indicated in ->ldp_attrs, in
936 * their natural order. After the last attribute, padding bytes are
937 * added to make ->lde_reclen a multiple of 8.
943 * Definitions of optional directory entry attributes formats.
945 * Individual attributes do not have their length encoded in a generic way. It
946 * is assumed that consumer of an attribute knows its format. This means that
947 * it is impossible to skip over an unknown attribute, except by skipping over all
948 * remaining attributes (by using ->lde_reclen), which is not too
949 * constraining, because new server versions will append new attributes at
950 * the end of an entry.
954 * Fid directory attribute: a fid of an object referenced by the entry. This
955 * will be almost always requested by the client and supplied by the server.
957 * Aligned to 8 bytes.
959 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
964 * Aligned to 2 bytes.
975 #define IFTODT(type) (((type) & S_IFMT) >> IFSHIFT)
978 #define DTTOIF(dirtype) ((dirtype) << IFSHIFT)
982 __u64 ldp_hash_start
;
986 struct lu_dirent ldp_entries
[0];
989 enum lu_dirpage_flags
{
991 * dirpage contains no entry.
995 * last entry's lde_hash equals ldp_hash_end.
1000 static inline struct lu_dirent
*lu_dirent_start(struct lu_dirpage
*dp
)
1002 if (le32_to_cpu(dp
->ldp_flags
) & LDF_EMPTY
)
1005 return dp
->ldp_entries
;
1008 static inline struct lu_dirent
*lu_dirent_next(struct lu_dirent
*ent
)
1010 struct lu_dirent
*next
;
1012 if (le16_to_cpu(ent
->lde_reclen
) != 0)
1013 next
= ((void *)ent
) + le16_to_cpu(ent
->lde_reclen
);
1020 static inline int lu_dirent_calc_size(int namelen
, __u16 attr
)
1024 if (attr
& LUDA_TYPE
) {
1025 const unsigned align
= sizeof(struct luda_type
) - 1;
1027 size
= (sizeof(struct lu_dirent
) + namelen
+ align
) & ~align
;
1028 size
+= sizeof(struct luda_type
);
1030 size
= sizeof(struct lu_dirent
) + namelen
;
1033 return (size
+ 7) & ~7;
1036 static inline int lu_dirent_size(const struct lu_dirent
*ent
)
1038 if (le16_to_cpu(ent
->lde_reclen
) == 0) {
1039 return lu_dirent_calc_size(le16_to_cpu(ent
->lde_namelen
),
1040 le32_to_cpu(ent
->lde_attrs
));
1042 return le16_to_cpu(ent
->lde_reclen
);
1045 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
1048 * MDS_READPAGE page size
1050 * This is the directory page size packed in MDS_READPAGE RPC.
1051 * It's different than PAGE_SIZE because the client needs to
1052 * access the struct lu_dirpage header packed at the beginning of
1053 * the "page" and without this there isn't any way to know find the
1054 * lu_dirpage header is if client and server PAGE_SIZE differ.
1056 #define LU_PAGE_SHIFT 12
1057 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
1058 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
1060 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
1064 struct lustre_handle
{
1068 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
1070 static inline bool lustre_handle_is_used(const struct lustre_handle
*lh
)
1072 return lh
->cookie
!= 0ull;
1075 static inline bool lustre_handle_equal(const struct lustre_handle
*lh1
,
1076 const struct lustre_handle
*lh2
)
1078 return lh1
->cookie
== lh2
->cookie
;
1081 static inline void lustre_handle_copy(struct lustre_handle
*tgt
,
1082 const struct lustre_handle
*src
)
1084 tgt
->cookie
= src
->cookie
;
1087 /* flags for lm_flags */
1088 #define MSGHDR_AT_SUPPORT 0x1
1089 #define MSGHDR_CKSUM_INCOMPAT18 0x2
1091 #define lustre_msg lustre_msg_v2
1092 /* we depend on this structure to be 8-byte aligned */
1093 /* this type is only endian-adjusted in lustre_unpack_msg() */
1094 struct lustre_msg_v2
{
1103 __u32 lm_buflens
[0];
1106 /* without gss, ptlrpc_body is put at the first buffer. */
1107 #define PTLRPC_NUM_VERSIONS 4
1108 #define JOBSTATS_JOBID_SIZE 32 /* 32 bytes string */
1109 struct ptlrpc_body_v3
{
1110 struct lustre_handle pb_handle
;
1117 __u64 pb_last_committed
;
1122 __u32 pb_timeout
; /* for req, the deadline, for rep, the service est */
1123 __u32 pb_service_time
; /* for rep, actual service time */
1126 /* VBR: pre-versions */
1127 __u64 pb_pre_versions
[PTLRPC_NUM_VERSIONS
];
1128 /* padding for future needs */
1129 __u64 pb_padding
[4];
1130 char pb_jobid
[JOBSTATS_JOBID_SIZE
];
1133 #define ptlrpc_body ptlrpc_body_v3
1135 struct ptlrpc_body_v2
{
1136 struct lustre_handle pb_handle
;
1143 __u64 pb_last_committed
;
1148 __u32 pb_timeout
; /* for req, the deadline, for rep, the service est */
1149 __u32 pb_service_time
; /* for rep, actual service time, also used for
1150 * net_latency of req
1154 /* VBR: pre-versions */
1155 __u64 pb_pre_versions
[PTLRPC_NUM_VERSIONS
];
1156 /* padding for future needs */
1157 __u64 pb_padding
[4];
1160 void lustre_swab_ptlrpc_body(struct ptlrpc_body
*pb
);
1162 /* message body offset for lustre_msg_v2 */
1163 /* ptlrpc body offset in all request/reply messages */
1164 #define MSG_PTLRPC_BODY_OFF 0
1166 /* normal request/reply message record offset */
1167 #define REQ_REC_OFF 1
1168 #define REPLY_REC_OFF 1
1170 /* ldlm request message body offset */
1171 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
1172 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
1174 /* ldlm intent lock message body offset */
1175 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
1176 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
1178 /* ldlm reply message body offset */
1179 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
1180 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
1182 /** only use in req->rq_{req,rep}_swab_mask */
1183 #define MSG_PTLRPC_HEADER_OFF 31
1185 /* Flags that are operation-specific go in the top 16 bits. */
1186 #define MSG_OP_FLAG_MASK 0xffff0000
1187 #define MSG_OP_FLAG_SHIFT 16
1189 /* Flags that apply to all requests are in the bottom 16 bits */
1190 #define MSG_GEN_FLAG_MASK 0x0000ffff
1191 #define MSG_LAST_REPLAY 0x0001
1192 #define MSG_RESENT 0x0002
1193 #define MSG_REPLAY 0x0004
1194 /* #define MSG_AT_SUPPORT 0x0008
1195 * This was used in early prototypes of adaptive timeouts, and while there
1196 * shouldn't be any users of that code there also isn't a need for using this
1197 * bits. Defer usage until at least 1.10 to avoid potential conflict.
1199 #define MSG_DELAY_REPLAY 0x0010
1200 #define MSG_VERSION_REPLAY 0x0020
1201 #define MSG_REQ_REPLAY_DONE 0x0040
1202 #define MSG_LOCK_REPLAY_DONE 0x0080
1205 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
1208 #define MSG_CONNECT_RECOVERING 0x00000001
1209 #define MSG_CONNECT_RECONNECT 0x00000002
1210 #define MSG_CONNECT_REPLAYABLE 0x00000004
1211 /*#define MSG_CONNECT_PEER 0x8 */
1212 #define MSG_CONNECT_LIBCLIENT 0x00000010
1213 #define MSG_CONNECT_INITIAL 0x00000020
1214 #define MSG_CONNECT_ASYNC 0x00000040
1215 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
1216 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
1219 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
1220 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
1221 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
1222 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
1223 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
1224 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
1225 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
1226 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
1227 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
1228 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
1229 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
1230 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
1231 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
1232 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
1233 *We do not support JOIN FILE
1234 *anymore, reserve this flags
1235 *just for preventing such bit
1238 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
1239 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
1240 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
1241 * in production. Removed in
1242 * 2.9. Keep this flag to
1245 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
1246 * never used in production.
1247 * Removed in 2.9. Keep this
1248 * flag to avoid reuse
1250 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
1251 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
1252 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
1253 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
1254 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
1255 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
1256 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
1257 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
1258 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
1259 #define OBD_CONNECT_REAL 0x8000000ULL /*real connection */
1260 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
1261 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
1262 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
1263 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
1264 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
1265 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
1266 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
1267 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
1268 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
1269 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
1270 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
1273 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
1274 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
1275 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
1276 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
1277 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
1278 * RPC error properly
1280 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
1281 * finer space reservation
1283 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
1284 * policy and 2.x server
1286 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
1287 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
1288 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
1289 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
1290 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
1291 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* flock deadlock detection */
1292 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/*create stripe disposition*/
1293 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
1298 * Please DO NOT add flag values here before first ensuring that this same
1299 * flag value is not in use on some other branch. Please clear any such
1300 * changes with senior engineers before starting to use a new flag. Then,
1301 * submit a small patch against EVERY branch that ONLY adds the new flag,
1302 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
1303 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
1304 * can be approved and landed easily to reserve the flag for future use.
1307 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
1308 * connection. It is a temporary bug fix for Imperative Recovery interop
1309 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
1310 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644.
1312 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
1314 #define OCD_HAS_FLAG(ocd, flg) \
1315 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
1317 /* Features required for this version of the client to work with server */
1318 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
1321 #define OBD_OCD_VERSION(major, minor, patch, fix) (((major)<<24) + \
1323 ((patch)<<8) + (fix))
1324 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
1325 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
1326 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
1327 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
1329 /* This structure is used for both request and reply.
1331 * If we eventually have separate connect data for different types, which we
1332 * almost certainly will, then perhaps we stick a union in here.
1334 struct obd_connect_data_v1
{
1335 __u64 ocd_connect_flags
; /* OBD_CONNECT_* per above */
1336 __u32 ocd_version
; /* lustre release version number */
1337 __u32 ocd_grant
; /* initial cache grant amount (bytes) */
1338 __u32 ocd_index
; /* LOV index to connect to */
1339 __u32 ocd_brw_size
; /* Maximum BRW size in bytes, must be 2^n */
1340 __u64 ocd_ibits_known
; /* inode bits this client understands */
1341 __u8 ocd_blocksize
; /* log2 of the backend filesystem blocksize */
1342 __u8 ocd_inodespace
; /* log2 of the per-inode space consumption */
1343 __u16 ocd_grant_extent
; /* per-extent grant overhead, in 1K blocks */
1344 __u32 ocd_unused
; /* also fix lustre_swab_connect */
1345 __u64 ocd_transno
; /* first transno from client to be replayed */
1346 __u32 ocd_group
; /* MDS group on OST */
1347 __u32 ocd_cksum_types
; /* supported checksum algorithms */
1348 __u32 ocd_max_easize
; /* How big LOV EA can be on MDS */
1349 __u32 ocd_instance
; /* also fix lustre_swab_connect */
1350 __u64 ocd_maxbytes
; /* Maximum stripe size in bytes */
1353 struct obd_connect_data
{
1354 __u64 ocd_connect_flags
; /* OBD_CONNECT_* per above */
1355 __u32 ocd_version
; /* lustre release version number */
1356 __u32 ocd_grant
; /* initial cache grant amount (bytes) */
1357 __u32 ocd_index
; /* LOV index to connect to */
1358 __u32 ocd_brw_size
; /* Maximum BRW size in bytes */
1359 __u64 ocd_ibits_known
; /* inode bits this client understands */
1360 __u8 ocd_blocksize
; /* log2 of the backend filesystem blocksize */
1361 __u8 ocd_inodespace
; /* log2 of the per-inode space consumption */
1362 __u16 ocd_grant_extent
; /* per-extent grant overhead, in 1K blocks */
1363 __u32 ocd_unused
; /* also fix lustre_swab_connect */
1364 __u64 ocd_transno
; /* first transno from client to be replayed */
1365 __u32 ocd_group
; /* MDS group on OST */
1366 __u32 ocd_cksum_types
; /* supported checksum algorithms */
1367 __u32 ocd_max_easize
; /* How big LOV EA can be on MDS */
1368 __u32 ocd_instance
; /* instance # of this target */
1369 __u64 ocd_maxbytes
; /* Maximum stripe size in bytes */
1370 /* Fields after ocd_maxbytes are only accessible by the receiver
1371 * if the corresponding flag in ocd_connect_flags is set. Accessing
1372 * any field after ocd_maxbytes on the receiver without a valid flag
1373 * may result in out-of-bound memory access and kernel oops.
1375 __u64 padding1
; /* added 2.1.0. also fix lustre_swab_connect */
1376 __u64 padding2
; /* added 2.1.0. also fix lustre_swab_connect */
1377 __u64 padding3
; /* added 2.1.0. also fix lustre_swab_connect */
1378 __u64 padding4
; /* added 2.1.0. also fix lustre_swab_connect */
1379 __u64 padding5
; /* added 2.1.0. also fix lustre_swab_connect */
1380 __u64 padding6
; /* added 2.1.0. also fix lustre_swab_connect */
1381 __u64 padding7
; /* added 2.1.0. also fix lustre_swab_connect */
1382 __u64 padding8
; /* added 2.1.0. also fix lustre_swab_connect */
1383 __u64 padding9
; /* added 2.1.0. also fix lustre_swab_connect */
1384 __u64 paddingA
; /* added 2.1.0. also fix lustre_swab_connect */
1385 __u64 paddingB
; /* added 2.1.0. also fix lustre_swab_connect */
1386 __u64 paddingC
; /* added 2.1.0. also fix lustre_swab_connect */
1387 __u64 paddingD
; /* added 2.1.0. also fix lustre_swab_connect */
1388 __u64 paddingE
; /* added 2.1.0. also fix lustre_swab_connect */
1389 __u64 paddingF
; /* added 2.1.0. also fix lustre_swab_connect */
1393 * Please DO NOT use any fields here before first ensuring that this same
1394 * field is not in use on some other branch. Please clear any such changes
1395 * with senior engineers before starting to use a new field. Then, submit
1396 * a small patch against EVERY branch that ONLY adds the new field along with
1397 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
1398 * reserve the flag for future use.
1401 void lustre_swab_connect(struct obd_connect_data
*ocd
);
1404 * Supported checksum algorithms. Up to 32 checksum types are supported.
1405 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1406 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
1407 * algorithm and also the OBD_FL_CKSUM* flags.
1410 OBD_CKSUM_CRC32
= 0x00000001,
1411 OBD_CKSUM_ADLER
= 0x00000002,
1412 OBD_CKSUM_CRC32C
= 0x00000004,
1416 * OST requests: OBDO & OBD request records
1421 OST_REPLY
= 0, /* reply ? */
1437 OST_QUOTACHECK
= 18,
1439 OST_QUOTA_ADJUST_QUNIT
= 20, /* not used since 2.4 */
1442 #define OST_FIRST_OPC OST_REPLY
1445 OBD_FL_INLINEDATA
= 0x00000001,
1446 OBD_FL_OBDMDEXISTS
= 0x00000002,
1447 OBD_FL_DELORPHAN
= 0x00000004, /* if set in o_flags delete orphans */
1448 OBD_FL_NORPC
= 0x00000008, /* set in o_flags do in OSC not OST */
1449 OBD_FL_IDONLY
= 0x00000010, /* set in o_flags only adjust obj id*/
1450 OBD_FL_RECREATE_OBJS
= 0x00000020, /* recreate missing obj */
1451 OBD_FL_DEBUG_CHECK
= 0x00000040, /* echo client/server debug check */
1452 OBD_FL_NO_USRQUOTA
= 0x00000100, /* the object's owner is over quota */
1453 OBD_FL_NO_GRPQUOTA
= 0x00000200, /* the object's group is over quota */
1454 OBD_FL_CREATE_CROW
= 0x00000400, /* object should be create on write */
1455 OBD_FL_SRVLOCK
= 0x00000800, /* delegate DLM locking to server */
1456 OBD_FL_CKSUM_CRC32
= 0x00001000, /* CRC32 checksum type */
1457 OBD_FL_CKSUM_ADLER
= 0x00002000, /* ADLER checksum type */
1458 OBD_FL_CKSUM_CRC32C
= 0x00004000, /* CRC32C checksum type */
1459 OBD_FL_CKSUM_RSVD2
= 0x00008000, /* for future cksum types */
1460 OBD_FL_CKSUM_RSVD3
= 0x00010000, /* for future cksum types */
1461 OBD_FL_SHRINK_GRANT
= 0x00020000, /* object shrink the grant */
1462 OBD_FL_MMAP
= 0x00040000, /* object is mmapped on the client.
1463 * XXX: obsoleted - reserved for old
1464 * clients prior than 2.2
1466 OBD_FL_RECOV_RESEND
= 0x00080000, /* recoverable resent */
1467 OBD_FL_NOSPC_BLK
= 0x00100000, /* no more block space on OST */
1468 OBD_FL_FLUSH
= 0x00200000, /* flush pages on the OST */
1469 OBD_FL_SHORT_IO
= 0x00400000, /* short io request */
1471 /* Note that while these checksum values are currently separate bits,
1472 * in 2.x we can actually allow all values from 1-31 if we wanted.
1474 OBD_FL_CKSUM_ALL
= OBD_FL_CKSUM_CRC32
| OBD_FL_CKSUM_ADLER
|
1475 OBD_FL_CKSUM_CRC32C
,
1477 /* mask for local-only flag, which won't be sent over network */
1478 OBD_FL_LOCAL_MASK
= 0xF0000000,
1482 * All LOV EA magics should have the same postfix, if some new version
1483 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1484 * Lustre, even though the old version system does not recognizes such
1485 * new magic, it still can distinguish the corrupted cases by checking
1486 * the magic's postfix.
1488 #define LOV_MAGIC_MAGIC 0x0BD0
1489 #define LOV_MAGIC_MASK 0xFFFF
1491 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1492 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1493 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1494 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1495 #define LOV_MAGIC LOV_MAGIC_V1
1498 * magic for fully defined striping
1499 * the idea is that we should have different magics for striping "hints"
1500 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1501 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1502 * we can't just change it w/o long way preparation, but we still need a
1503 * mechanism to allow LOD to differentiate hint versus ready striping.
1504 * so, at the moment we do a trick: MDT knows what to expect from request
1505 * depending on the case (replay uses ready striping, non-replay req uses
1506 * hints), so MDT replaces magic with appropriate one and now LOD can
1507 * easily understand what's inside -bzzz
1509 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
1510 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
1512 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
1513 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
1514 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
1515 #define LOV_PATTERN_CMOBD 0x200
1517 #define LOV_PATTERN_F_MASK 0xffff0000
1518 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
1520 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1521 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1523 #define lov_ost_data lov_ost_data_v1
1524 struct lov_ost_data_v1
{ /* per-stripe data structure (little-endian)*/
1525 struct ost_id l_ost_oi
; /* OST object ID */
1526 __u32 l_ost_gen
; /* generation of this l_ost_idx */
1527 __u32 l_ost_idx
; /* OST index in LOV (lov_tgt_desc->tgts) */
1530 #define lov_mds_md lov_mds_md_v1
1531 struct lov_mds_md_v1
{ /* LOV EA mds/wire data (little-endian) */
1532 __u32 lmm_magic
; /* magic number = LOV_MAGIC_V1 */
1533 __u32 lmm_pattern
; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1534 struct ost_id lmm_oi
; /* LOV object ID */
1535 __u32 lmm_stripe_size
; /* size of stripe in bytes */
1536 /* lmm_stripe_count used to be __u32 */
1537 __u16 lmm_stripe_count
; /* num stripes in use for this object */
1538 __u16 lmm_layout_gen
; /* layout generation number */
1539 struct lov_ost_data_v1 lmm_objects
[0]; /* per-stripe data */
1543 * Sigh, because pre-2.4 uses
1544 * struct lov_mds_md_v1 {
1546 * __u64 lmm_object_id;
1547 * __u64 lmm_object_seq;
1550 * to identify the LOV(MDT) object, and lmm_object_seq will
1551 * be normal_fid, which make it hard to combine these conversion
1552 * to ostid_to FID. so we will do lmm_oi/fid conversion separately
1554 * We can tell the lmm_oi by this way,
1555 * 1.8: lmm_object_id = {inode}, lmm_object_gr = 0
1556 * 2.1: lmm_object_id = {oid < 128k}, lmm_object_seq = FID_SEQ_NORMAL
1557 * 2.4: lmm_oi.f_seq = FID_SEQ_NORMAL, lmm_oi.f_oid = {oid < 128k},
1560 * But currently lmm_oi/lsm_oi does not have any "real" usages,
1561 * except for printing some information, and the user can always
1562 * get the real FID from LMA, besides this multiple case check might
1563 * make swab more complicate. So we will keep using id/seq for lmm_oi.
1566 static inline void fid_to_lmm_oi(const struct lu_fid
*fid
,
1569 oi
->oi
.oi_id
= fid_oid(fid
);
1570 oi
->oi
.oi_seq
= fid_seq(fid
);
1573 static inline void lmm_oi_set_seq(struct ost_id
*oi
, __u64 seq
)
1575 oi
->oi
.oi_seq
= seq
;
1578 static inline void lmm_oi_set_id(struct ost_id
*oi
, __u64 oid
)
1583 static inline __u64
lmm_oi_id(const struct ost_id
*oi
)
1585 return oi
->oi
.oi_id
;
1588 static inline __u64
lmm_oi_seq(const struct ost_id
*oi
)
1590 return oi
->oi
.oi_seq
;
1593 static inline void lmm_oi_le_to_cpu(struct ost_id
*dst_oi
,
1594 const struct ost_id
*src_oi
)
1596 dst_oi
->oi
.oi_id
= le64_to_cpu(src_oi
->oi
.oi_id
);
1597 dst_oi
->oi
.oi_seq
= le64_to_cpu(src_oi
->oi
.oi_seq
);
1600 static inline void lmm_oi_cpu_to_le(struct ost_id
*dst_oi
,
1601 const struct ost_id
*src_oi
)
1603 dst_oi
->oi
.oi_id
= cpu_to_le64(src_oi
->oi
.oi_id
);
1604 dst_oi
->oi
.oi_seq
= cpu_to_le64(src_oi
->oi
.oi_seq
);
1607 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
1609 #define MAX_MD_SIZE \
1610 (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1611 #define MIN_MD_SIZE \
1612 (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1614 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1615 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1616 #define XATTR_USER_PREFIX "user."
1617 #define XATTR_TRUSTED_PREFIX "trusted."
1618 #define XATTR_SECURITY_PREFIX "security."
1619 #define XATTR_LUSTRE_PREFIX "lustre."
1621 #define XATTR_NAME_LOV "trusted.lov"
1622 #define XATTR_NAME_LMA "trusted.lma"
1623 #define XATTR_NAME_LMV "trusted.lmv"
1624 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1625 #define XATTR_NAME_LINK "trusted.link"
1626 #define XATTR_NAME_FID "trusted.fid"
1627 #define XATTR_NAME_VERSION "trusted.version"
1628 #define XATTR_NAME_SOM "trusted.som"
1629 #define XATTR_NAME_HSM "trusted.hsm"
1630 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_namespace"
1632 struct lov_mds_md_v3
{ /* LOV EA mds/wire data (little-endian) */
1633 __u32 lmm_magic
; /* magic number = LOV_MAGIC_V3 */
1634 __u32 lmm_pattern
; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1635 struct ost_id lmm_oi
; /* LOV object ID */
1636 __u32 lmm_stripe_size
; /* size of stripe in bytes */
1637 /* lmm_stripe_count used to be __u32 */
1638 __u16 lmm_stripe_count
; /* num stripes in use for this object */
1639 __u16 lmm_layout_gen
; /* layout generation number */
1640 char lmm_pool_name
[LOV_MAXPOOLNAME
]; /* must be 32bit aligned */
1641 struct lov_ost_data_v1 lmm_objects
[0]; /* per-stripe data */
1644 static inline __u32
lov_mds_md_size(__u16 stripes
, __u32 lmm_magic
)
1646 if (lmm_magic
== LOV_MAGIC_V3
)
1647 return sizeof(struct lov_mds_md_v3
) +
1648 stripes
* sizeof(struct lov_ost_data_v1
);
1650 return sizeof(struct lov_mds_md_v1
) +
1651 stripes
* sizeof(struct lov_ost_data_v1
);
1655 lov_mds_md_max_stripe_count(size_t buf_size
, __u32 lmm_magic
)
1657 switch (lmm_magic
) {
1658 case LOV_MAGIC_V1
: {
1659 struct lov_mds_md_v1 lmm
;
1661 if (buf_size
< sizeof(lmm
))
1664 return (buf_size
- sizeof(lmm
)) / sizeof(lmm
.lmm_objects
[0]);
1666 case LOV_MAGIC_V3
: {
1667 struct lov_mds_md_v3 lmm
;
1669 if (buf_size
< sizeof(lmm
))
1672 return (buf_size
- sizeof(lmm
)) / sizeof(lmm
.lmm_objects
[0]);
1679 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1680 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1681 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1682 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1683 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1684 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1685 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1686 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1687 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1688 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1689 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1690 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1691 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1692 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1693 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
1694 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1695 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1696 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1697 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1698 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1699 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1700 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1701 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
1702 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1703 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1704 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1705 /* ->mds if epoch opens or closes
1707 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1708 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1709 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1710 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1711 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1713 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1714 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1715 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1716 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1718 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1719 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1720 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1721 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1722 /* OBD_MD_FLRMTPERM (0x0000010000000000ULL) remote perm, obsolete */
1723 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1724 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1725 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1726 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1727 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1728 * under lock; for xattr
1729 * requests means the
1730 * client holds the lock
1732 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1734 /* OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) lfs lsetfacl, obsolete */
1735 /* OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) lfs lgetfacl, obsolete */
1736 /* OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) lfs rsetfacl, obsolete */
1737 /* OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) lfs rgetfacl, obsolete */
1739 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1740 #define OBD_MD_FLRELEASED (0x0020000000000000ULL) /* file released */
1742 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1744 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1745 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1746 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1747 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1748 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1750 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1752 /* don't forget obdo_fid which is way down at the bottom so it can
1753 * come after the definition of llog_cookie
1758 HSS_CLEARMASK
= 0x02,
1759 HSS_ARCHIVE_ID
= 0x04,
1762 struct hsm_state_set
{
1764 __u32 hss_archive_id
;
1766 __u64 hss_clearmask
;
1769 void lustre_swab_hsm_user_state(struct hsm_user_state
*hus
);
1770 void lustre_swab_hsm_state_set(struct hsm_state_set
*hss
);
1772 void lustre_swab_obd_statfs(struct obd_statfs
*os
);
1774 /* ost_body.data values for OST_BRW */
1776 #define OBD_BRW_READ 0x01
1777 #define OBD_BRW_WRITE 0x02
1778 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1779 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1780 * transfer and is not accounted in
1783 #define OBD_BRW_CHECK 0x10
1784 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1785 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1786 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1787 #define OBD_BRW_NOQUOTA 0x100
1788 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1789 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1790 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1791 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1792 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1793 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1794 * that the client is running low on
1795 * space for unstable pages; asking
1796 * it to sync quickly
1799 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
1801 #define OST_MIN_PRECREATE 32
1802 #define OST_MAX_PRECREATE 20000
1805 struct ost_id ioo_oid
; /* object ID, if multi-obj BRW */
1806 __u32 ioo_max_brw
; /* low 16 bits were o_mode before 2.4,
1807 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1808 * high 16 bits in 2.4 and later
1810 __u32 ioo_bufcnt
; /* number of niobufs for this object */
1813 #define IOOBJ_MAX_BRW_BITS 16
1814 #define IOOBJ_TYPE_MASK ((1U << IOOBJ_MAX_BRW_BITS) - 1)
1815 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1816 #define ioobj_max_brw_set(ioo, num) \
1817 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1819 void lustre_swab_obd_ioobj(struct obd_ioobj
*ioo
);
1821 /* multiple of 8 bytes => can array */
1822 struct niobuf_remote
{
1828 void lustre_swab_niobuf_remote(struct niobuf_remote
*nbr
);
1830 /* lock value block communicated between the filter and llite */
1832 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1833 * negative, i.e. because ((MASK + rc) & MASK) != MASK.
1835 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1836 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1837 #define OST_LVB_IS_ERR(blocks) \
1838 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1839 #define OST_LVB_SET_ERR(blocks, rc) \
1840 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1841 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1851 void lustre_swab_ost_lvb_v1(struct ost_lvb_v1
*lvb
);
1865 void lustre_swab_ost_lvb(struct ost_lvb
*lvb
);
1868 * lquota data structures
1871 /* The lquota_id structure is an union of all the possible identifier types that
1872 * can be used with quota, this includes:
1875 * - a FID which can be used for per-directory quota in the future
1878 struct lu_fid qid_fid
; /* FID for per-directory quota */
1879 __u64 qid_uid
; /* user identifier */
1880 __u64 qid_gid
; /* group identifier */
1883 /* quotactl management */
1884 struct obd_quotactl
{
1886 __u32 qc_type
; /* see Q_* flag below */
1889 struct obd_dqinfo qc_dqinfo
;
1890 struct obd_dqblk qc_dqblk
;
1893 void lustre_swab_obd_quotactl(struct obd_quotactl
*q
);
1895 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1896 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1897 #define Q_GETOINFO 0x800102 /* get obd quota info */
1898 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1899 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1901 #define Q_COPY(out, in, member) (out)->member = (in)->member
1903 #define QCTL_COPY(out, in) \
1905 Q_COPY(out, in, qc_cmd); \
1906 Q_COPY(out, in, qc_type); \
1907 Q_COPY(out, in, qc_id); \
1908 Q_COPY(out, in, qc_stat); \
1909 Q_COPY(out, in, qc_dqinfo); \
1910 Q_COPY(out, in, qc_dqblk); \
1913 /* Data structures associated with the quota locks */
1915 /* Glimpse descriptor used for the index & per-ID quota locks */
1916 struct ldlm_gl_lquota_desc
{
1917 union lquota_id gl_id
; /* quota ID subject to the glimpse */
1918 __u64 gl_flags
; /* see LQUOTA_FL* below */
1919 __u64 gl_ver
; /* new index version */
1920 __u64 gl_hardlimit
; /* new hardlimit or qunit value */
1921 __u64 gl_softlimit
; /* new softlimit */
1926 /* quota glimpse flags */
1927 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1929 /* LVB used with quota (global and per-ID) locks */
1931 __u64 lvb_flags
; /* see LQUOTA_FL* above */
1932 __u64 lvb_id_may_rel
; /* space that might be released later */
1933 __u64 lvb_id_rel
; /* space released by the slave for this ID */
1934 __u64 lvb_id_qunit
; /* current qunit value */
1938 void lustre_swab_lquota_lvb(struct lquota_lvb
*lvb
);
1946 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1955 MDS_GETATTR_NAME
= 34,
1960 MDS_DISCONNECT
= 39,
1966 MDS_DONE_WRITING
= 45,
1968 MDS_QUOTACHECK
= 47,
1971 MDS_SETXATTR
= 50, /* obsolete, now it's MDS_REINT op */
1975 MDS_HSM_STATE_GET
= 54,
1976 MDS_HSM_STATE_SET
= 55,
1977 MDS_HSM_ACTION
= 56,
1978 MDS_HSM_PROGRESS
= 57,
1979 MDS_HSM_REQUEST
= 58,
1980 MDS_HSM_CT_REGISTER
= 59,
1981 MDS_HSM_CT_UNREGISTER
= 60,
1982 MDS_SWAP_LAYOUTS
= 61,
1986 #define MDS_FIRST_OPC MDS_GETATTR
1992 enum mdt_reint_cmd
{
2005 void lustre_swab_generic_32s(__u32
*val
);
2007 /* the disposition of the intent outlines what was executed */
2008 #define DISP_IT_EXECD 0x00000001
2009 #define DISP_LOOKUP_EXECD 0x00000002
2010 #define DISP_LOOKUP_NEG 0x00000004
2011 #define DISP_LOOKUP_POS 0x00000008
2012 #define DISP_OPEN_CREATE 0x00000010
2013 #define DISP_OPEN_OPEN 0x00000020
2014 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
2015 #define DISP_ENQ_OPEN_REF 0x00800000
2016 #define DISP_ENQ_CREATE_REF 0x01000000
2017 #define DISP_OPEN_LOCK 0x02000000
2018 #define DISP_OPEN_LEASE 0x04000000
2019 #define DISP_OPEN_STRIPE 0x08000000
2021 /* INODE LOCK PARTS */
2022 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
2023 * was used to protect permission (mode,
2024 * owner, group etc) before 2.4.
2026 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
2027 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
2028 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
2030 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
2031 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
2032 * Because for remote directories(in DNE), these locks will be granted by
2033 * different MDTs(different ldlm namespace).
2035 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
2036 * For Remote directory, the master MDT, where the remote directory is, will
2037 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
2038 * will grant LOOKUP_LOCK.
2040 #define MDS_INODELOCK_PERM 0x000010
2041 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
2043 #define MDS_INODELOCK_MAXSHIFT 5
2044 /* This FULL lock is useful to take on unlink sort of operations */
2045 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
2047 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
2048 * but was moved into name[1] along with the OID to avoid consuming the
2049 * name[2,3] fields that need to be used for the quota id (also a FID).
2052 LUSTRE_RES_ID_SEQ_OFF
= 0,
2053 LUSTRE_RES_ID_VER_OID_OFF
= 1,
2054 LUSTRE_RES_ID_WAS_VER_OFF
= 2, /* see note above */
2055 LUSTRE_RES_ID_QUOTA_SEQ_OFF
= 2,
2056 LUSTRE_RES_ID_QUOTA_VER_OID_OFF
= 3,
2057 LUSTRE_RES_ID_HSH_OFF
= 3
2060 #define MDS_STATUS_CONN 1
2061 #define MDS_STATUS_LOV 2
2063 /* mdt_thread_info.mti_flags. */
2065 /* The flag indicates Size-on-MDS attributes are changed. */
2066 MF_SOM_CHANGE
= (1 << 0),
2067 /* Flags indicates an epoch opens or closes. */
2068 MF_EPOCH_OPEN
= (1 << 1),
2069 MF_EPOCH_CLOSE
= (1 << 2),
2070 MF_MDC_CANCEL_FID1
= (1 << 3),
2071 MF_MDC_CANCEL_FID2
= (1 << 4),
2072 MF_MDC_CANCEL_FID3
= (1 << 5),
2073 MF_MDC_CANCEL_FID4
= (1 << 6),
2074 /* There is a pending attribute update. */
2075 MF_SOM_AU
= (1 << 7),
2076 /* Cancel OST locks while getattr OST attributes. */
2077 MF_GETATTR_LOCK
= (1 << 8),
2078 MF_GET_MDT_IDX
= (1 << 9),
2081 #define MF_SOM_LOCAL_FLAGS (MF_SOM_CHANGE | MF_EPOCH_OPEN | MF_EPOCH_CLOSE)
2083 #define LUSTRE_BFLAG_UNCOMMITTED_WRITES 0x1
2085 /* these should be identical to their EXT4_*_FL counterparts, they are
2086 * redefined here only to avoid dragging in fs/ext4/ext4.h
2088 #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */
2089 #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */
2090 #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */
2091 #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */
2092 #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
2094 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
2095 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
2096 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
2097 * the S_* flags are kernel-internal values that change between kernel
2098 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
2099 * See b=16526 for a full history.
2101 static inline int ll_ext_to_inode_flags(int flags
)
2103 return (((flags
& LUSTRE_SYNC_FL
) ? S_SYNC
: 0) |
2104 ((flags
& LUSTRE_NOATIME_FL
) ? S_NOATIME
: 0) |
2105 ((flags
& LUSTRE_APPEND_FL
) ? S_APPEND
: 0) |
2106 ((flags
& LUSTRE_DIRSYNC_FL
) ? S_DIRSYNC
: 0) |
2107 ((flags
& LUSTRE_IMMUTABLE_FL
) ? S_IMMUTABLE
: 0));
2110 static inline int ll_inode_to_ext_flags(int iflags
)
2112 return (((iflags
& S_SYNC
) ? LUSTRE_SYNC_FL
: 0) |
2113 ((iflags
& S_NOATIME
) ? LUSTRE_NOATIME_FL
: 0) |
2114 ((iflags
& S_APPEND
) ? LUSTRE_APPEND_FL
: 0) |
2115 ((iflags
& S_DIRSYNC
) ? LUSTRE_DIRSYNC_FL
: 0) |
2116 ((iflags
& S_IMMUTABLE
) ? LUSTRE_IMMUTABLE_FL
: 0));
2119 /* 64 possible states */
2120 enum md_transient_state
{
2121 MS_RESTORE
= (1 << 0), /* restore is running */
2127 struct lustre_handle handle
;
2129 __u64 size
; /* Offset, in the case of MDS_READPAGE */
2133 __u64 blocks
; /* XID, in the case of MDS_READPAGE */
2135 __u64 t_state
; /* transient file state defined in
2136 * enum md_transient_state
2137 * was "ino" until 2.4.0
2145 __u32 flags
; /* from vfs for pin/unpin, LUSTRE_BFLAG close */
2147 __u32 nlink
; /* #bytes to read in the case of MDS_READPAGE */
2148 __u32 unused2
; /* was "generation" until 2.4.0 */
2153 __u32 max_cookiesize
;
2154 __u32 uid_h
; /* high 32-bits of uid, for FUID */
2155 __u32 gid_h
; /* high 32-bits of gid, for FUID */
2156 __u32 padding_5
; /* also fix lustre_swab_mdt_body */
2164 void lustre_swab_mdt_body(struct mdt_body
*b
);
2166 struct mdt_ioepoch
{
2167 struct lustre_handle handle
;
2173 void lustre_swab_mdt_ioepoch(struct mdt_ioepoch
*b
);
2175 /* permissions for md_perm.mp_perm */
2177 CFS_SETUID_PERM
= 0x01,
2178 CFS_SETGID_PERM
= 0x02,
2179 CFS_SETGRP_PERM
= 0x04,
2182 struct mdt_rec_setattr
{
2192 __u32 sa_padding_1_h
;
2193 struct lu_fid sa_fid
;
2202 __u32 sa_attr_flags
;
2204 __u32 sa_bias
; /* some operation flags */
2210 void lustre_swab_mdt_rec_setattr(struct mdt_rec_setattr
*sa
);
2213 * Attribute flags used in mdt_rec_setattr::sa_valid.
2214 * The kernel's #defines for ATTR_* should not be used over the network
2215 * since the client and MDS may run different kernels (see bug 13828)
2216 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
2218 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
2219 #define MDS_ATTR_UID 0x2ULL /* = 2 */
2220 #define MDS_ATTR_GID 0x4ULL /* = 4 */
2221 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
2222 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
2223 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
2224 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
2225 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
2226 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
2227 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
2228 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
2229 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
2230 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
2231 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
2232 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path,
2235 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
2237 #define MDS_FMODE_CLOSED 00000000
2238 #define MDS_FMODE_EXEC 00000004
2239 /* IO Epoch is opened on a closed file. */
2240 #define MDS_FMODE_EPOCH 01000000
2241 /* IO Epoch is opened on a file truncate. */
2242 #define MDS_FMODE_TRUNC 02000000
2243 /* Size-on-MDS Attribute Update is pending. */
2244 #define MDS_FMODE_SOM 04000000
2246 #define MDS_OPEN_CREATED 00000010
2247 #define MDS_OPEN_CROSS 00000020
2249 #define MDS_OPEN_CREAT 00000100
2250 #define MDS_OPEN_EXCL 00000200
2251 #define MDS_OPEN_TRUNC 00001000
2252 #define MDS_OPEN_APPEND 00002000
2253 #define MDS_OPEN_SYNC 00010000
2254 #define MDS_OPEN_DIRECTORY 00200000
2256 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
2257 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
2258 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
2259 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
2260 * We do not support JOIN FILE
2261 * anymore, reserve this flags
2262 * just for preventing such bit
2266 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
2267 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
2268 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
2269 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
2270 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
2272 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
2274 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
2275 * delegation, succeed if it's not
2276 * being opened with conflict mode.
2278 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
2281 MDS_CHECK_SPLIT
= 1 << 0,
2282 MDS_CROSS_REF
= 1 << 1,
2283 MDS_VTX_BYPASS
= 1 << 2,
2284 MDS_PERM_BYPASS
= 1 << 3,
2286 MDS_QUOTA_IGNORE
= 1 << 5,
2287 MDS_CLOSE_CLEANUP
= 1 << 6,
2288 MDS_KEEP_ORPHAN
= 1 << 7,
2289 MDS_RECOV_OPEN
= 1 << 8,
2290 MDS_DATA_MODIFIED
= 1 << 9,
2291 MDS_CREATE_VOLATILE
= 1 << 10,
2292 MDS_OWNEROVERRIDE
= 1 << 11,
2293 MDS_HSM_RELEASE
= 1 << 12,
2294 MDS_RENAME_MIGRATE
= BIT(13),
2297 /* instance of mdt_reint_rec */
2298 struct mdt_rec_create
{
2306 __u32 cr_suppgid1_h
;
2308 __u32 cr_suppgid2_h
;
2309 struct lu_fid cr_fid1
;
2310 struct lu_fid cr_fid2
;
2311 struct lustre_handle cr_old_handle
; /* handle in case of open replay */
2315 __u64 cr_padding_1
; /* rr_blocks */
2318 /* use of helpers set/get_mrc_cr_flags() is needed to access
2319 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2320 * extend cr_flags size without breaking 1.8 compat
2322 __u32 cr_flags_l
; /* for use with open, low 32 bits */
2323 __u32 cr_flags_h
; /* for use with open, high 32 bits */
2324 __u32 cr_umask
; /* umask for create */
2325 __u32 cr_padding_4
; /* rr_padding_4 */
2328 static inline void set_mrc_cr_flags(struct mdt_rec_create
*mrc
, __u64 flags
)
2330 mrc
->cr_flags_l
= (__u32
)(flags
& 0xFFFFFFFFUll
);
2331 mrc
->cr_flags_h
= (__u32
)(flags
>> 32);
2334 static inline __u64
get_mrc_cr_flags(struct mdt_rec_create
*mrc
)
2336 return ((__u64
)(mrc
->cr_flags_l
) | ((__u64
)mrc
->cr_flags_h
<< 32));
2339 /* instance of mdt_reint_rec */
2340 struct mdt_rec_link
{
2348 __u32 lk_suppgid1_h
;
2350 __u32 lk_suppgid2_h
;
2351 struct lu_fid lk_fid1
;
2352 struct lu_fid lk_fid2
;
2354 __u64 lk_padding_1
; /* rr_atime */
2355 __u64 lk_padding_2
; /* rr_ctime */
2356 __u64 lk_padding_3
; /* rr_size */
2357 __u64 lk_padding_4
; /* rr_blocks */
2359 __u32 lk_padding_5
; /* rr_mode */
2360 __u32 lk_padding_6
; /* rr_flags */
2361 __u32 lk_padding_7
; /* rr_padding_2 */
2362 __u32 lk_padding_8
; /* rr_padding_3 */
2363 __u32 lk_padding_9
; /* rr_padding_4 */
2366 /* instance of mdt_reint_rec */
2367 struct mdt_rec_unlink
{
2375 __u32 ul_suppgid1_h
;
2377 __u32 ul_suppgid2_h
;
2378 struct lu_fid ul_fid1
;
2379 struct lu_fid ul_fid2
;
2381 __u64 ul_padding_2
; /* rr_atime */
2382 __u64 ul_padding_3
; /* rr_ctime */
2383 __u64 ul_padding_4
; /* rr_size */
2384 __u64 ul_padding_5
; /* rr_blocks */
2387 __u32 ul_padding_6
; /* rr_flags */
2388 __u32 ul_padding_7
; /* rr_padding_2 */
2389 __u32 ul_padding_8
; /* rr_padding_3 */
2390 __u32 ul_padding_9
; /* rr_padding_4 */
2393 /* instance of mdt_reint_rec */
2394 struct mdt_rec_rename
{
2402 __u32 rn_suppgid1_h
;
2404 __u32 rn_suppgid2_h
;
2405 struct lu_fid rn_fid1
;
2406 struct lu_fid rn_fid2
;
2408 __u64 rn_padding_1
; /* rr_atime */
2409 __u64 rn_padding_2
; /* rr_ctime */
2410 __u64 rn_padding_3
; /* rr_size */
2411 __u64 rn_padding_4
; /* rr_blocks */
2412 __u32 rn_bias
; /* some operation flags */
2413 __u32 rn_mode
; /* cross-ref rename has mode */
2414 __u32 rn_padding_5
; /* rr_flags */
2415 __u32 rn_padding_6
; /* rr_padding_2 */
2416 __u32 rn_padding_7
; /* rr_padding_3 */
2417 __u32 rn_padding_8
; /* rr_padding_4 */
2420 /* instance of mdt_reint_rec */
2421 struct mdt_rec_setxattr
{
2429 __u32 sx_suppgid1_h
;
2431 __u32 sx_suppgid2_h
;
2432 struct lu_fid sx_fid
;
2433 __u64 sx_padding_1
; /* These three are rr_fid2 */
2438 __u64 sx_padding_5
; /* rr_ctime */
2439 __u64 sx_padding_6
; /* rr_size */
2440 __u64 sx_padding_7
; /* rr_blocks */
2443 __u32 sx_padding_8
; /* rr_flags */
2444 __u32 sx_padding_9
; /* rr_padding_2 */
2445 __u32 sx_padding_10
; /* rr_padding_3 */
2446 __u32 sx_padding_11
; /* rr_padding_4 */
2450 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2451 * Do NOT change the size of various members, otherwise the value
2452 * will be broken in lustre_swab_mdt_rec_reint().
2454 * If you add new members in other mdt_reint_xxx structures and need to use the
2455 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2457 struct mdt_rec_reint
{
2465 __u32 rr_suppgid1_h
;
2467 __u32 rr_suppgid2_h
;
2468 struct lu_fid rr_fid1
;
2469 struct lu_fid rr_fid2
;
2480 __u32 rr_padding_4
; /* also fix lustre_swab_mdt_rec_reint */
2483 void lustre_swab_mdt_rec_reint(struct mdt_rec_reint
*rr
);
2485 /* lmv structures */
2487 __u32 ld_tgt_count
; /* how many MDS's */
2488 __u32 ld_active_tgt_count
; /* how many active */
2489 __u32 ld_default_stripe_count
; /* how many objects are used */
2490 __u32 ld_pattern
; /* default hash pattern */
2491 __u64 ld_default_hash_size
;
2492 __u64 ld_padding_1
; /* also fix lustre_swab_lmv_desc */
2493 __u32 ld_padding_2
; /* also fix lustre_swab_lmv_desc */
2494 __u32 ld_qos_maxage
; /* in second */
2495 __u32 ld_padding_3
; /* also fix lustre_swab_lmv_desc */
2496 __u32 ld_padding_4
; /* also fix lustre_swab_lmv_desc */
2497 struct obd_uuid ld_uuid
;
2500 /* LMV layout EA, and it will be stored both in master and slave object */
2501 struct lmv_mds_md_v1
{
2503 __u32 lmv_stripe_count
;
2504 __u32 lmv_master_mdt_index
; /* On master object, it is master
2505 * MDT index, on slave object, it
2506 * is stripe index of the slave obj
2508 __u32 lmv_hash_type
; /* dir stripe policy, i.e. indicate
2509 * which hash function to be used,
2510 * Note: only lower 16 bits is being
2511 * used for now. Higher 16 bits will
2512 * be used to mark the object status,
2513 * for example migrating or dead.
2515 __u32 lmv_layout_version
; /* Used for directory restriping */
2517 struct lu_fid lmv_master_fid
; /* The FID of the master object, which
2518 * is the namespace-visible dir FID
2520 char lmv_pool_name
[LOV_MAXPOOLNAME
]; /* pool name */
2521 struct lu_fid lmv_stripe_fids
[0]; /* FIDs for each stripe */
2524 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2525 #define LMV_MAGIC LMV_MAGIC_V1
2527 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2528 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2531 *Right now only the lower part(0-16bits) of lmv_hash_type is being used,
2532 * and the higher part will be the flag to indicate the status of object,
2533 * for example the object is being migrated. And the hash function
2534 * might be interpreted differently with different flags.
2536 enum lmv_hash_type
{
2537 LMV_HASH_TYPE_ALL_CHARS
= 1,
2538 LMV_HASH_TYPE_FNV_1A_64
= 2,
2541 #define LMV_HASH_TYPE_MASK 0x0000ffff
2543 #define LMV_HASH_FLAG_MIGRATION 0x80000000
2544 #define LMV_HASH_FLAG_DEAD 0x40000000
2546 #define LMV_HASH_NAME_ALL_CHARS "all_char"
2547 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
2550 * The FNV-1a hash algorithm is as follows:
2551 * hash = FNV_offset_basis
2552 * for each octet_of_data to be hashed
2553 * hash = hash XOR octet_of_data
2554 * hash = hash × FNV_prime
2556 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2558 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2559 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2561 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2562 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2563 static inline __u64
lustre_hash_fnv_1a_64(const void *buf
, size_t size
)
2565 __u64 hash
= LUSTRE_FNV_1A_64_OFFSET_BIAS
;
2566 const unsigned char *p
= buf
;
2569 for (i
= 0; i
< size
; i
++) {
2571 hash
*= LUSTRE_FNV_1A_64_PRIME
;
2579 struct lmv_mds_md_v1 lmv_md_v1
;
2580 struct lmv_user_md lmv_user_md
;
2583 void lustre_swab_lmv_mds_md(union lmv_mds_md
*lmm
);
2585 static inline ssize_t
lmv_mds_md_size(int stripe_count
, unsigned int lmm_magic
)
2587 ssize_t len
= -EINVAL
;
2589 switch (lmm_magic
) {
2590 case LMV_MAGIC_V1
: {
2591 struct lmv_mds_md_v1
*lmm1
;
2593 len
= sizeof(*lmm1
);
2594 len
+= stripe_count
* sizeof(lmm1
->lmv_stripe_fids
[0]);
2602 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md
*lmm
)
2604 switch (le32_to_cpu(lmm
->lmv_magic
)) {
2606 return le32_to_cpu(lmm
->lmv_md_v1
.lmv_stripe_count
);
2607 case LMV_USER_MAGIC
:
2608 return le32_to_cpu(lmm
->lmv_user_md
.lum_stripe_count
);
2614 static inline int lmv_mds_md_stripe_count_set(union lmv_mds_md
*lmm
,
2615 unsigned int stripe_count
)
2619 switch (le32_to_cpu(lmm
->lmv_magic
)) {
2621 lmm
->lmv_md_v1
.lmv_stripe_count
= cpu_to_le32(stripe_count
);
2623 case LMV_USER_MAGIC
:
2624 lmm
->lmv_user_md
.lum_stripe_count
= cpu_to_le32(stripe_count
);
2637 FLD_FIRST_OPC
= FLD_QUERY
2643 SEQ_FIRST_OPC
= SEQ_QUERY
2647 SEQ_ALLOC_SUPER
= 0,
2658 * LOV data structures
2661 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2662 /* The size of the buffer the lov/mdc reserves for the
2663 * array of UUIDs returned by the MDS. With the current
2664 * protocol, this will limit the max number of OSTs per LOV
2667 #define LOV_DESC_MAGIC 0xB0CCDE5C
2668 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2669 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2671 /* LOV settings descriptor (should only contain static info) */
2673 __u32 ld_tgt_count
; /* how many OBD's */
2674 __u32 ld_active_tgt_count
; /* how many active */
2675 __u32 ld_default_stripe_count
; /* how many objects are used */
2676 __u32 ld_pattern
; /* default PATTERN_RAID0 */
2677 __u64 ld_default_stripe_size
; /* in bytes */
2678 __u64 ld_default_stripe_offset
; /* in bytes */
2679 __u32 ld_padding_0
; /* unused */
2680 __u32 ld_qos_maxage
; /* in second */
2681 __u32 ld_padding_1
; /* also fix lustre_swab_lov_desc */
2682 __u32 ld_padding_2
; /* also fix lustre_swab_lov_desc */
2683 struct obd_uuid ld_uuid
;
2686 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2688 void lustre_swab_lov_desc(struct lov_desc
*ld
);
2693 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2698 LDLM_BL_CALLBACK
= 104,
2699 LDLM_CP_CALLBACK
= 105,
2700 LDLM_GL_CALLBACK
= 106,
2701 LDLM_SET_INFO
= 107,
2704 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2706 #define RES_NAME_SIZE 4
2707 struct ldlm_res_id
{
2708 __u64 name
[RES_NAME_SIZE
];
2711 #define DLDLMRES "[%#llx:%#llx:%#llx].%llx"
2712 #define PLDLMRES(res) (res)->lr_name.name[0], (res)->lr_name.name[1], \
2713 (res)->lr_name.name[2], (res)->lr_name.name[3]
2715 static inline bool ldlm_res_eq(const struct ldlm_res_id
*res0
,
2716 const struct ldlm_res_id
*res1
)
2718 return !memcmp(res0
, res1
, sizeof(*res0
));
2735 #define LCK_MODE_NUM 8
2745 #define LDLM_MIN_TYPE LDLM_PLAIN
2747 struct ldlm_extent
{
2753 #define LDLM_GID_ANY ((__u64)-1)
2755 static inline int ldlm_extent_overlap(const struct ldlm_extent
*ex1
,
2756 const struct ldlm_extent
*ex2
)
2758 return (ex1
->start
<= ex2
->end
) && (ex2
->start
<= ex1
->end
);
2761 /* check if @ex1 contains @ex2 */
2762 static inline int ldlm_extent_contain(const struct ldlm_extent
*ex1
,
2763 const struct ldlm_extent
*ex2
)
2765 return (ex1
->start
<= ex2
->start
) && (ex1
->end
>= ex2
->end
);
2768 struct ldlm_inodebits
{
2772 struct ldlm_flock_wire
{
2780 /* it's important that the fields of the ldlm_extent structure match
2781 * the first fields of the ldlm_flock structure because there is only
2782 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2783 * this ever changes we will need to swab the union differently based
2784 * on the resource type.
2788 struct ldlm_extent l_extent
;
2789 struct ldlm_flock_wire l_flock
;
2790 struct ldlm_inodebits l_inodebits
;
2791 } ldlm_wire_policy_data_t
;
2793 union ldlm_gl_desc
{
2794 struct ldlm_gl_lquota_desc lquota_desc
;
2797 void lustre_swab_gl_desc(union ldlm_gl_desc
*);
2799 struct ldlm_intent
{
2803 void lustre_swab_ldlm_intent(struct ldlm_intent
*i
);
2805 struct ldlm_resource_desc
{
2806 enum ldlm_type lr_type
;
2807 __u32 lr_padding
; /* also fix lustre_swab_ldlm_resource_desc */
2808 struct ldlm_res_id lr_name
;
2811 struct ldlm_lock_desc
{
2812 struct ldlm_resource_desc l_resource
;
2813 enum ldlm_mode l_req_mode
;
2814 enum ldlm_mode l_granted_mode
;
2815 ldlm_wire_policy_data_t l_policy_data
;
2818 #define LDLM_LOCKREQ_HANDLES 2
2819 #define LDLM_ENQUEUE_CANCEL_OFF 1
2821 struct ldlm_request
{
2824 struct ldlm_lock_desc lock_desc
;
2825 struct lustre_handle lock_handle
[LDLM_LOCKREQ_HANDLES
];
2828 void lustre_swab_ldlm_request(struct ldlm_request
*rq
);
2830 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
2831 * Otherwise, 2 are available.
2833 #define ldlm_request_bufsize(count, type) \
2835 int _avail = LDLM_LOCKREQ_HANDLES; \
2836 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
2837 sizeof(struct ldlm_request) + \
2838 (count > _avail ? count - _avail : 0) * \
2839 sizeof(struct lustre_handle); \
2844 __u32 lock_padding
; /* also fix lustre_swab_ldlm_reply */
2845 struct ldlm_lock_desc lock_desc
;
2846 struct lustre_handle lock_handle
;
2847 __u64 lock_policy_res1
;
2848 __u64 lock_policy_res2
;
2851 void lustre_swab_ldlm_reply(struct ldlm_reply
*r
);
2853 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2854 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2857 * Opcodes for mountconf (mgs and mgc)
2862 MGS_EXCEPTION
, /* node died, etc. */
2863 MGS_TARGET_REG
, /* whenever target starts up */
2869 #define MGS_FIRST_OPC MGS_CONNECT
2871 #define MGS_PARAM_MAXLEN 1024
2872 #define KEY_SET_INFO "set_info"
2874 struct mgs_send_param
{
2875 char mgs_param
[MGS_PARAM_MAXLEN
];
2878 /* We pass this info to the MGS so it can write config logs */
2879 #define MTI_NAME_MAXLEN 64
2880 #define MTI_PARAM_MAXLEN 4096
2881 #define MTI_NIDS_MAX 32
2882 struct mgs_target_info
{
2883 __u32 mti_lustre_ver
;
2884 __u32 mti_stripe_index
;
2885 __u32 mti_config_ver
;
2887 __u32 mti_nid_count
;
2888 __u32 mti_instance
; /* Running instance of target */
2889 char mti_fsname
[MTI_NAME_MAXLEN
];
2890 char mti_svname
[MTI_NAME_MAXLEN
];
2891 char mti_uuid
[sizeof(struct obd_uuid
)];
2892 __u64 mti_nids
[MTI_NIDS_MAX
]; /* host nids (lnet_nid_t)*/
2893 char mti_params
[MTI_PARAM_MAXLEN
];
2896 void lustre_swab_mgs_target_info(struct mgs_target_info
*oinfo
);
2898 struct mgs_nidtbl_entry
{
2899 __u64 mne_version
; /* table version of this entry */
2900 __u32 mne_instance
; /* target instance # */
2901 __u32 mne_index
; /* target index */
2902 __u32 mne_length
; /* length of this entry - by bytes */
2903 __u8 mne_type
; /* target type LDD_F_SV_TYPE_OST/MDT */
2904 __u8 mne_nid_type
; /* type of nid(mbz). for ipv6. */
2905 __u8 mne_nid_size
; /* size of each NID, by bytes */
2906 __u8 mne_nid_count
; /* # of NIDs in buffer */
2908 lnet_nid_t nids
[0]; /* variable size buffer for NIDs. */
2912 void lustre_swab_mgs_nidtbl_entry(struct mgs_nidtbl_entry
*oinfo
);
2914 struct mgs_config_body
{
2915 char mcb_name
[MTI_NAME_MAXLEN
]; /* logname */
2916 __u64 mcb_offset
; /* next index of config log to request */
2917 __u16 mcb_type
; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2919 __u8 mcb_bits
; /* bits unit size of config log */
2920 __u32 mcb_units
; /* # of units for bulk transfer */
2923 void lustre_swab_mgs_config_body(struct mgs_config_body
*body
);
2925 struct mgs_config_res
{
2926 __u64 mcr_offset
; /* index of last config log */
2927 __u64 mcr_size
; /* size of the log */
2930 void lustre_swab_mgs_config_res(struct mgs_config_res
*body
);
2932 /* Config marker flags (in config log) */
2933 #define CM_START 0x01
2935 #define CM_SKIP 0x04
2936 #define CM_UPGRADE146 0x08
2937 #define CM_EXCLUDE 0x10
2938 #define CM_START_SKIP (CM_START | CM_SKIP)
2941 __u32 cm_step
; /* aka config version */
2943 __u32 cm_vers
; /* lustre release version number */
2944 __u32 cm_padding
; /* 64 bit align */
2945 __s64 cm_createtime
; /*when this record was first created */
2946 __s64 cm_canceltime
; /*when this record is no longer valid*/
2947 char cm_tgtname
[MTI_NAME_MAXLEN
];
2948 char cm_comment
[MTI_NAME_MAXLEN
];
2951 void lustre_swab_cfg_marker(struct cfg_marker
*marker
, int swab
, int size
);
2954 * Opcodes for multiple servers.
2964 #define OBD_FIRST_OPC OBD_PING
2966 /* catalog of log objects */
2968 /** Identifier for a single log object */
2970 struct ost_id lgl_oi
;
2974 /** Records written to the CATALOGS list */
2975 #define CATLIST "CATALOGS"
2977 struct llog_logid lci_logid
;
2983 /* Log data record types - there is no specific reason that these need to
2984 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2986 #define LLOG_OP_MAGIC 0x10600000
2987 #define LLOG_OP_MASK 0xfff00000
2990 LLOG_PAD_MAGIC
= LLOG_OP_MAGIC
| 0x00000,
2991 OST_SZ_REC
= LLOG_OP_MAGIC
| 0x00f00,
2992 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2993 MDS_UNLINK_REC
= LLOG_OP_MAGIC
| 0x10000 | (MDS_REINT
<< 8) |
2994 REINT_UNLINK
, /* obsolete after 2.5.0 */
2995 MDS_UNLINK64_REC
= LLOG_OP_MAGIC
| 0x90000 | (MDS_REINT
<< 8) |
2997 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2998 MDS_SETATTR64_REC
= LLOG_OP_MAGIC
| 0x90000 | (MDS_REINT
<< 8) |
3000 OBD_CFG_REC
= LLOG_OP_MAGIC
| 0x20000,
3001 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
3002 LLOG_GEN_REC
= LLOG_OP_MAGIC
| 0x40000,
3003 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
3004 CHANGELOG_REC
= LLOG_OP_MAGIC
| 0x60000,
3005 CHANGELOG_USER_REC
= LLOG_OP_MAGIC
| 0x70000,
3006 HSM_AGENT_REC
= LLOG_OP_MAGIC
| 0x80000,
3007 LLOG_HDR_MAGIC
= LLOG_OP_MAGIC
| 0x45539,
3008 LLOG_LOGID_MAGIC
= LLOG_OP_MAGIC
| 0x4553b,
3011 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
3012 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
3014 /** Log record header - stored in little endian order.
3015 * Each record must start with this struct, end with a llog_rec_tail,
3016 * and be a multiple of 256 bits in size.
3018 struct llog_rec_hdr
{
3025 struct llog_rec_tail
{
3030 /* Where data follow just after header */
3031 #define REC_DATA(ptr) \
3032 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
3034 #define REC_DATA_LEN(rec) \
3035 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
3036 sizeof(struct llog_rec_tail))
3038 struct llog_logid_rec
{
3039 struct llog_rec_hdr lid_hdr
;
3040 struct llog_logid lid_id
;
3044 struct llog_rec_tail lid_tail
;
3047 struct llog_unlink_rec
{
3048 struct llog_rec_hdr lur_hdr
;
3052 struct llog_rec_tail lur_tail
;
3055 struct llog_unlink64_rec
{
3056 struct llog_rec_hdr lur_hdr
;
3057 struct lu_fid lur_fid
;
3058 __u32 lur_count
; /* to destroy the lost precreated */
3062 struct llog_rec_tail lur_tail
;
3065 struct llog_setattr64_rec
{
3066 struct llog_rec_hdr lsr_hdr
;
3067 struct ost_id lsr_oi
;
3073 struct llog_rec_tail lsr_tail
;
3076 struct llog_size_change_rec
{
3077 struct llog_rec_hdr lsc_hdr
;
3078 struct ll_fid lsc_fid
;
3083 struct llog_rec_tail lsc_tail
;
3086 /* changelog llog name, needed by client replicators */
3087 #define CHANGELOG_CATALOG "changelog_catalog"
3089 struct changelog_setinfo
{
3094 /** changelog record */
3095 struct llog_changelog_rec
{
3096 struct llog_rec_hdr cr_hdr
;
3097 struct changelog_rec cr
;
3098 struct llog_rec_tail cr_tail
; /**< for_sizezof_only */
3101 struct llog_changelog_ext_rec
{
3102 struct llog_rec_hdr cr_hdr
;
3103 struct changelog_ext_rec cr
;
3104 struct llog_rec_tail cr_tail
; /**< for_sizezof_only */
3107 struct llog_changelog_user_rec
{
3108 struct llog_rec_hdr cur_hdr
;
3112 struct llog_rec_tail cur_tail
;
3115 enum agent_req_status
{
3123 static inline const char *agent_req_status2name(const enum agent_req_status ars
)
3141 static inline bool agent_req_in_final_state(enum agent_req_status ars
)
3143 return ((ars
== ARS_SUCCEED
) || (ars
== ARS_FAILED
) ||
3144 (ars
== ARS_CANCELED
));
3147 struct llog_agent_req_rec
{
3148 struct llog_rec_hdr arr_hdr
; /**< record header */
3149 __u32 arr_status
; /**< status of the request */
3153 __u32 arr_archive_id
; /**< backend archive number */
3154 __u64 arr_flags
; /**< req flags */
3155 __u64 arr_compound_id
;/**< compound cookie */
3156 __u64 arr_req_create
; /**< req. creation time */
3157 __u64 arr_req_change
; /**< req. status change time */
3158 struct hsm_action_item arr_hai
; /**< req. to the agent */
3159 struct llog_rec_tail arr_tail
; /**< record tail for_sizezof_only */
3162 /* Old llog gen for compatibility */
3168 struct llog_gen_rec
{
3169 struct llog_rec_hdr lgr_hdr
;
3170 struct llog_gen lgr_gen
;
3174 struct llog_rec_tail lgr_tail
;
3177 /* On-disk header structure of each log object, stored in little endian order */
3178 #define LLOG_CHUNK_SIZE 8192
3179 #define LLOG_HEADER_SIZE (96)
3180 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
3182 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
3184 /* flags for the logs */
3186 LLOG_F_ZAP_WHEN_EMPTY
= 0x1,
3187 LLOG_F_IS_CAT
= 0x2,
3188 LLOG_F_IS_PLAIN
= 0x4,
3191 struct llog_log_hdr
{
3192 struct llog_rec_hdr llh_hdr
;
3193 __s64 llh_timestamp
;
3195 __u32 llh_bitmap_offset
;
3199 /* for a catalog the first plain slot is next to it */
3200 struct obd_uuid llh_tgtuuid
;
3201 __u32 llh_reserved
[LLOG_HEADER_SIZE
/sizeof(__u32
) - 23];
3202 __u32 llh_bitmap
[LLOG_BITMAP_BYTES
/sizeof(__u32
)];
3203 struct llog_rec_tail llh_tail
;
3206 #define LLOG_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
3207 llh->llh_bitmap_offset - \
3208 sizeof(llh->llh_tail)) * 8)
3210 /** log cookies are used to reference a specific log file and a record
3213 struct llog_cookie
{
3214 struct llog_logid lgc_lgl
;
3220 /** llog protocol */
3221 enum llogd_rpc_ops
{
3222 LLOG_ORIGIN_HANDLE_CREATE
= 501,
3223 LLOG_ORIGIN_HANDLE_NEXT_BLOCK
= 502,
3224 LLOG_ORIGIN_HANDLE_READ_HEADER
= 503,
3225 LLOG_ORIGIN_HANDLE_WRITE_REC
= 504,
3226 LLOG_ORIGIN_HANDLE_CLOSE
= 505,
3227 LLOG_ORIGIN_CONNECT
= 506,
3228 LLOG_CATINFO
= 507, /* deprecated */
3229 LLOG_ORIGIN_HANDLE_PREV_BLOCK
= 508,
3230 LLOG_ORIGIN_HANDLE_DESTROY
= 509, /* for destroy llog object*/
3232 LLOG_FIRST_OPC
= LLOG_ORIGIN_HANDLE_CREATE
3236 struct llog_logid lgd_logid
;
3238 __u32 lgd_llh_flags
;
3240 __u32 lgd_saved_index
;
3242 __u64 lgd_cur_offset
;
3245 struct llogd_conn_body
{
3246 struct llog_gen lgdc_gen
;
3247 struct llog_logid lgdc_logid
;
3248 __u32 lgdc_ctxt_idx
;
3251 /* Note: 64-bit types are 64-bit aligned in structure */
3253 __u64 o_valid
; /* hot fields in this obdo */
3256 __u64 o_size
; /* o_size-o_blocks == ost_lvb */
3260 __u64 o_blocks
; /* brw: cli sent cached bytes */
3263 /* 32-bit fields start here: keep an even number of them via padding */
3264 __u32 o_blksize
; /* optimal IO blocksize */
3265 __u32 o_mode
; /* brw: cli sent cache remain */
3269 __u32 o_nlink
; /* brw: checksum */
3271 __u32 o_misc
; /* brw: o_dropped */
3273 __u64 o_ioepoch
; /* epoch in ost writes */
3274 __u32 o_stripe_idx
; /* holds stripe idx */
3276 struct lustre_handle o_handle
; /* brw: lock handle to prolong locks
3278 struct llog_cookie o_lcookie
; /* destroy: unlink cookie from MDS
3283 __u64 o_data_version
; /* getattr: sum of iversion for
3285 * brw: grant space consumed on
3286 * the client for the write
3293 #define o_dirty o_blocks
3294 #define o_undirty o_mode
3295 #define o_dropped o_misc
3296 #define o_cksum o_nlink
3297 #define o_grant_used o_data_version
3299 static inline void lustre_set_wire_obdo(const struct obd_connect_data
*ocd
,
3301 const struct obdo
*lobdo
)
3304 wobdo
->o_flags
&= ~OBD_FL_LOCAL_MASK
;
3308 if (unlikely(!(ocd
->ocd_connect_flags
& OBD_CONNECT_FID
)) &&
3309 fid_seq_is_echo(ostid_seq(&lobdo
->o_oi
))) {
3310 /* Currently OBD_FL_OSTID will only be used when 2.4 echo
3311 * client communicate with pre-2.4 server
3313 wobdo
->o_oi
.oi
.oi_id
= fid_oid(&lobdo
->o_oi
.oi_fid
);
3314 wobdo
->o_oi
.oi
.oi_seq
= fid_seq(&lobdo
->o_oi
.oi_fid
);
3318 static inline void lustre_get_wire_obdo(const struct obd_connect_data
*ocd
,
3320 const struct obdo
*wobdo
)
3322 __u32 local_flags
= 0;
3324 if (lobdo
->o_valid
& OBD_MD_FLFLAGS
)
3325 local_flags
= lobdo
->o_flags
& OBD_FL_LOCAL_MASK
;
3328 if (local_flags
!= 0) {
3329 lobdo
->o_valid
|= OBD_MD_FLFLAGS
;
3330 lobdo
->o_flags
&= ~OBD_FL_LOCAL_MASK
;
3331 lobdo
->o_flags
|= local_flags
;
3336 if (unlikely(!(ocd
->ocd_connect_flags
& OBD_CONNECT_FID
)) &&
3337 fid_seq_is_echo(wobdo
->o_oi
.oi
.oi_seq
)) {
3339 lobdo
->o_oi
.oi_fid
.f_seq
= wobdo
->o_oi
.oi
.oi_seq
;
3340 lobdo
->o_oi
.oi_fid
.f_oid
= wobdo
->o_oi
.oi
.oi_id
;
3341 lobdo
->o_oi
.oi_fid
.f_ver
= 0;
3345 /* request structure for OST's */
3350 /* Key for FIEMAP to be used in get_info calls */
3351 struct ll_fiemap_info_key
{
3354 struct ll_user_fiemap fiemap
;
3357 void lustre_swab_ost_body(struct ost_body
*b
);
3358 void lustre_swab_ost_last_id(__u64
*id
);
3359 void lustre_swab_fiemap(struct ll_user_fiemap
*fiemap
);
3361 void lustre_swab_lov_user_md_v1(struct lov_user_md_v1
*lum
);
3362 void lustre_swab_lov_user_md_v3(struct lov_user_md_v3
*lum
);
3363 void lustre_swab_lov_user_md_objects(struct lov_user_ost_data
*lod
,
3365 void lustre_swab_lov_mds_md(struct lov_mds_md
*lmm
);
3368 void lustre_swab_llogd_body(struct llogd_body
*d
);
3369 void lustre_swab_llog_hdr(struct llog_log_hdr
*h
);
3370 void lustre_swab_llogd_conn_body(struct llogd_conn_body
*d
);
3371 void lustre_swab_llog_rec(struct llog_rec_hdr
*rec
);
3374 void lustre_swab_lustre_cfg(struct lustre_cfg
*lcfg
);
3376 /* Functions for dumping PTLRPC fields */
3377 void dump_rniobuf(struct niobuf_remote
*rnb
);
3378 void dump_ioo(struct obd_ioobj
*nb
);
3379 void dump_ost_body(struct ost_body
*ob
);
3380 void dump_rcs(__u32
*rc
);
3382 /* security opcodes */
3385 SEC_CTX_INIT_CONT
= 802,
3388 SEC_FIRST_OPC
= SEC_CTX_INIT
3392 * capa related definitions
3394 #define CAPA_HMAC_MAX_LEN 64
3395 #define CAPA_HMAC_KEY_MAX_LEN 56
3397 /* NB take care when changing the sequence of elements this struct,
3398 * because the offset info is used in find_capa()
3400 struct lustre_capa
{
3401 struct lu_fid lc_fid
; /** fid */
3402 __u64 lc_opc
; /** operations allowed */
3403 __u64 lc_uid
; /** file owner */
3404 __u64 lc_gid
; /** file group */
3405 __u32 lc_flags
; /** HMAC algorithm & flags */
3406 __u32 lc_keyid
; /** key# used for the capability */
3407 __u32 lc_timeout
; /** capa timeout value (sec) */
3408 /* FIXME: y2038 time_t overflow: */
3409 __u32 lc_expiry
; /** expiry time (sec) */
3410 __u8 lc_hmac
[CAPA_HMAC_MAX_LEN
]; /** HMAC */
3413 void lustre_swab_lustre_capa(struct lustre_capa
*c
);
3415 /** lustre_capa::lc_opc */
3417 CAPA_OPC_BODY_WRITE
= 1<<0, /**< write object data */
3418 CAPA_OPC_BODY_READ
= 1<<1, /**< read object data */
3419 CAPA_OPC_INDEX_LOOKUP
= 1<<2, /**< lookup object fid */
3420 CAPA_OPC_INDEX_INSERT
= 1<<3, /**< insert object fid */
3421 CAPA_OPC_INDEX_DELETE
= 1<<4, /**< delete object fid */
3422 CAPA_OPC_OSS_WRITE
= 1<<5, /**< write oss object data */
3423 CAPA_OPC_OSS_READ
= 1<<6, /**< read oss object data */
3424 CAPA_OPC_OSS_TRUNC
= 1<<7, /**< truncate oss object */
3425 CAPA_OPC_OSS_DESTROY
= 1<<8, /**< destroy oss object */
3426 CAPA_OPC_META_WRITE
= 1<<9, /**< write object meta data */
3427 CAPA_OPC_META_READ
= 1<<10, /**< read object meta data */
3430 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3431 #define CAPA_OPC_MDS_ONLY \
3432 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3433 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3434 #define CAPA_OPC_OSS_ONLY \
3435 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3436 CAPA_OPC_OSS_DESTROY)
3437 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3438 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3440 struct lustre_capa_key
{
3441 __u64 lk_seq
; /**< mds# */
3442 __u32 lk_keyid
; /**< key# */
3444 __u8 lk_key
[CAPA_HMAC_KEY_MAX_LEN
]; /**< key */
3447 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3448 #define LINK_EA_MAGIC 0x11EAF1DFUL
3449 struct link_ea_header
{
3452 __u64 leh_len
; /* total size */
3458 /** Hardlink data is name and parent fid.
3459 * Stored in this crazy struct for maximum packing and endian-neutrality
3461 struct link_ea_entry
{
3462 /** __u16 stored big-endian, unaligned */
3463 unsigned char lee_reclen
[2];
3464 unsigned char lee_parent_fid
[sizeof(struct lu_fid
)];
3468 /** fid2path request/reply structure */
3469 struct getinfo_fid2path
{
3470 struct lu_fid gf_fid
;
3477 void lustre_swab_fid2path(struct getinfo_fid2path
*gf
);
3480 LAYOUT_INTENT_ACCESS
= 0,
3481 LAYOUT_INTENT_READ
= 1,
3482 LAYOUT_INTENT_WRITE
= 2,
3483 LAYOUT_INTENT_GLIMPSE
= 3,
3484 LAYOUT_INTENT_TRUNC
= 4,
3485 LAYOUT_INTENT_RELEASE
= 5,
3486 LAYOUT_INTENT_RESTORE
= 6
3489 /* enqueue layout lock with intent */
3490 struct layout_intent
{
3491 __u32 li_opc
; /* intent operation for enqueue, read, write etc */
3497 void lustre_swab_layout_intent(struct layout_intent
*li
);
3500 * On the wire version of hsm_progress structure.
3502 * Contains the userspace hsm_progress and some internal fields.
3504 struct hsm_progress_kernel
{
3505 /* Field taken from struct hsm_progress */
3506 struct lu_fid hpk_fid
;
3508 struct hsm_extent hpk_extent
;
3510 __u16 hpk_errval
; /* positive val */
3512 /* Additional fields */
3513 __u64 hpk_data_version
;
3517 void lustre_swab_hsm_user_state(struct hsm_user_state
*hus
);
3518 void lustre_swab_hsm_current_action(struct hsm_current_action
*action
);
3519 void lustre_swab_hsm_progress_kernel(struct hsm_progress_kernel
*hpk
);
3520 void lustre_swab_hsm_user_state(struct hsm_user_state
*hus
);
3521 void lustre_swab_hsm_user_item(struct hsm_user_item
*hui
);
3522 void lustre_swab_hsm_request(struct hsm_request
*hr
);
3524 /** layout swap request structure
3525 * fid1 and fid2 are in mdt_body
3527 struct mdc_swap_layouts
{
3531 void lustre_swab_swap_layouts(struct mdc_swap_layouts
*msl
);
3534 struct lustre_handle cd_handle
;
3535 struct lu_fid cd_fid
;
3536 __u64 cd_data_version
;
3537 __u64 cd_reserved
[8];
3540 void lustre_swab_close_data(struct close_data
*data
);