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.sun.com/software/products/lustre/docs/GPLv2.pdf
20 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21 * CA 95054 USA or visit www.sun.com if you need additional information or
27 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
28 * Use is subject to license terms.
30 * Copyright (c) 2011, 2012, Intel Corporation.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre/lustre_idl.h
38 * Lustre wire protocol definitions.
41 /** \defgroup lustreidl lustreidl
43 * Lustre wire protocol definitions.
45 * ALL structs passing over the wire should be declared here. Structs
46 * that are used in interfaces with userspace should go in lustre_user.h.
48 * All structs being declared here should be built from simple fixed-size
49 * types (__u8, __u16, __u32, __u64) or be built from other types or
50 * structs also declared in this file. Similarly, all flags and magic
51 * values in those structs should also be declared here. This ensures
52 * that the Lustre wire protocol is not influenced by external dependencies.
54 * The only other acceptable items in this file are VERY SIMPLE accessor
55 * functions to avoid callers grubbing inside the structures, and the
56 * prototypes of the swabber functions for each struct. Nothing that
57 * depends on external functions or definitions should be in here.
59 * Structs must be properly aligned to put 64-bit values on an 8-byte
60 * boundary. Any structs being added here must also be added to
61 * utils/wirecheck.c and "make newwiretest" run to regenerate the
62 * utils/wiretest.c sources. This allows us to verify that wire structs
63 * have the proper alignment/size on all architectures.
65 * DO NOT CHANGE any of the structs, flags, values declared here and used
66 * in released Lustre versions. Some structs may have padding fields that
67 * can be used. Some structs might allow addition at the end (verify this
68 * in the code to ensure that new/old clients that see this larger struct
69 * do not fail, otherwise you need to implement protocol compatibility).
71 * We assume all nodes are either little-endian or big-endian, and we
72 * always send messages in the sender's native format. The receiver
73 * detects the message format by checking the 'magic' field of the message
74 * (see lustre_msg_swabbed() below).
76 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
77 * implemented either here, inline (trivial implementations) or in
78 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
79 * endian, in-place in the message buffer.
81 * A swabber takes a single pointer argument. The caller must already have
82 * verified that the length of the message buffer >= sizeof (type).
84 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
85 * may be defined that swabs just the variable part, after the caller has
86 * verified that the message buffer is large enough.
91 #ifndef _LUSTRE_IDL_H_
92 #define _LUSTRE_IDL_H_
94 #include "../../../include/linux/libcfs/libcfs.h"
96 /* Defn's shared with user-space. */
97 #include "lustre_user.h"
98 #include "lustre_errno.h"
103 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
104 * FOO_REPLY_PORTAL is for incoming replies on the FOO
105 * FOO_BULK_PORTAL is for incoming bulk on the FOO
108 #define CONNMGR_REQUEST_PORTAL 1
109 #define CONNMGR_REPLY_PORTAL 2
110 //#define OSC_REQUEST_PORTAL 3
111 #define OSC_REPLY_PORTAL 4
112 //#define OSC_BULK_PORTAL 5
113 #define OST_IO_PORTAL 6
114 #define OST_CREATE_PORTAL 7
115 #define OST_BULK_PORTAL 8
116 //#define MDC_REQUEST_PORTAL 9
117 #define MDC_REPLY_PORTAL 10
118 //#define MDC_BULK_PORTAL 11
119 #define MDS_REQUEST_PORTAL 12
120 //#define MDS_REPLY_PORTAL 13
121 #define MDS_BULK_PORTAL 14
122 #define LDLM_CB_REQUEST_PORTAL 15
123 #define LDLM_CB_REPLY_PORTAL 16
124 #define LDLM_CANCEL_REQUEST_PORTAL 17
125 #define LDLM_CANCEL_REPLY_PORTAL 18
126 //#define PTLBD_REQUEST_PORTAL 19
127 //#define PTLBD_REPLY_PORTAL 20
128 //#define PTLBD_BULK_PORTAL 21
129 #define MDS_SETATTR_PORTAL 22
130 #define MDS_READPAGE_PORTAL 23
131 #define MDS_MDS_PORTAL 24
133 #define MGC_REPLY_PORTAL 25
134 #define MGS_REQUEST_PORTAL 26
135 #define MGS_REPLY_PORTAL 27
136 #define OST_REQUEST_PORTAL 28
137 #define FLD_REQUEST_PORTAL 29
138 #define SEQ_METADATA_PORTAL 30
139 #define SEQ_DATA_PORTAL 31
140 #define SEQ_CONTROLLER_PORTAL 32
141 #define MGS_BULK_PORTAL 33
143 /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com, n8851@cray.com */
146 #define PTL_RPC_MSG_REQUEST 4711
147 #define PTL_RPC_MSG_ERR 4712
148 #define PTL_RPC_MSG_REPLY 4713
150 /* DON'T use swabbed values of MAGIC as magic! */
151 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
152 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
154 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
155 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
157 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
159 #define PTLRPC_MSG_VERSION 0x00000003
160 #define LUSTRE_VERSION_MASK 0xffff0000
161 #define LUSTRE_OBD_VERSION 0x00010000
162 #define LUSTRE_MDS_VERSION 0x00020000
163 #define LUSTRE_OST_VERSION 0x00030000
164 #define LUSTRE_DLM_VERSION 0x00040000
165 #define LUSTRE_LOG_VERSION 0x00050000
166 #define LUSTRE_MGS_VERSION 0x00060000
168 typedef __u32 mdsno_t
;
169 typedef __u64 seqno_t
;
170 typedef __u64 obd_id
;
171 typedef __u64 obd_seq
;
172 typedef __s64 obd_time
;
173 typedef __u64 obd_size
;
174 typedef __u64 obd_off
;
175 typedef __u64 obd_blocks
;
176 typedef __u64 obd_valid
;
177 typedef __u32 obd_blksize
;
178 typedef __u32 obd_mode
;
179 typedef __u32 obd_uid
;
180 typedef __u32 obd_gid
;
181 typedef __u32 obd_flag
;
182 typedef __u32 obd_count
;
185 * Describes a range of sequence, lsr_start is included but lsr_end is
187 * Same structure is used in fld module where lsr_index field holds mdt id
190 struct lu_seq_range
{
197 #define LU_SEQ_RANGE_MDT 0x0
198 #define LU_SEQ_RANGE_OST 0x1
199 #define LU_SEQ_RANGE_ANY 0x3
201 #define LU_SEQ_RANGE_MASK 0x3
203 static inline unsigned fld_range_type(const struct lu_seq_range
*range
)
205 return range
->lsr_flags
& LU_SEQ_RANGE_MASK
;
208 static inline int fld_range_is_ost(const struct lu_seq_range
*range
)
210 return fld_range_type(range
) == LU_SEQ_RANGE_OST
;
213 static inline int fld_range_is_mdt(const struct lu_seq_range
*range
)
215 return fld_range_type(range
) == LU_SEQ_RANGE_MDT
;
219 * This all range is only being used when fld client sends fld query request,
220 * but it does not know whether the seq is MDT or OST, so it will send req
221 * with ALL type, which means either seq type gotten from lookup can be
224 static inline unsigned fld_range_is_any(const struct lu_seq_range
*range
)
226 return fld_range_type(range
) == LU_SEQ_RANGE_ANY
;
229 static inline void fld_range_set_type(struct lu_seq_range
*range
,
232 range
->lsr_flags
|= flags
;
235 static inline void fld_range_set_mdt(struct lu_seq_range
*range
)
237 fld_range_set_type(range
, LU_SEQ_RANGE_MDT
);
240 static inline void fld_range_set_ost(struct lu_seq_range
*range
)
242 fld_range_set_type(range
, LU_SEQ_RANGE_OST
);
245 static inline void fld_range_set_any(struct lu_seq_range
*range
)
247 fld_range_set_type(range
, LU_SEQ_RANGE_ANY
);
251 * returns width of given range \a r
254 static inline __u64
range_space(const struct lu_seq_range
*range
)
256 return range
->lsr_end
- range
->lsr_start
;
260 * initialize range to zero
263 static inline void range_init(struct lu_seq_range
*range
)
265 memset(range
, 0, sizeof(*range
));
269 * check if given seq id \a s is within given range \a r
272 static inline int range_within(const struct lu_seq_range
*range
,
275 return s
>= range
->lsr_start
&& s
< range
->lsr_end
;
278 static inline int range_is_sane(const struct lu_seq_range
*range
)
280 return (range
->lsr_end
>= range
->lsr_start
);
283 static inline int range_is_zero(const struct lu_seq_range
*range
)
285 return (range
->lsr_start
== 0 && range
->lsr_end
== 0);
288 static inline int range_is_exhausted(const struct lu_seq_range
*range
)
291 return range_space(range
) == 0;
294 /* return 0 if two range have the same location */
295 static inline int range_compare_loc(const struct lu_seq_range
*r1
,
296 const struct lu_seq_range
*r2
)
298 return r1
->lsr_index
!= r2
->lsr_index
||
299 r1
->lsr_flags
!= r2
->lsr_flags
;
302 #define DRANGE "[%#16.16Lx-%#16.16Lx):%x:%s"
304 #define PRANGE(range) \
305 (range)->lsr_start, \
307 (range)->lsr_index, \
308 fld_range_is_mdt(range) ? "mdt" : "ost"
311 /** \defgroup lu_fid lu_fid
315 * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat.
316 * Deprecated since HSM and SOM attributes are now stored in separate on-disk
320 LMAC_HSM
= 0x00000001,
321 LMAC_SOM
= 0x00000002,
322 LMAC_NOT_IN_OI
= 0x00000004, /* the object does NOT need OI mapping */
323 LMAC_FID_ON_OST
= 0x00000008, /* For OST-object, its OI mapping is
324 * under /O/<seq>/d<x>. */
328 * Masks for all features that should be supported by a Lustre version to
329 * access a specific file.
330 * This information is stored in lustre_mdt_attrs::lma_incompat.
333 LMAI_RELEASED
= 0x00000001, /* file is released */
334 LMAI_AGENT
= 0x00000002, /* agent inode */
335 LMAI_REMOTE_PARENT
= 0x00000004, /* the parent of the object
336 is on the remote MDT */
338 #define LMA_INCOMPAT_SUPP (LMAI_AGENT | LMAI_REMOTE_PARENT)
340 extern void lustre_lma_swab(struct lustre_mdt_attrs
*lma
);
341 extern void lustre_lma_init(struct lustre_mdt_attrs
*lma
,
342 const struct lu_fid
*fid
, __u32 incompat
);
344 * SOM on-disk attributes stored in a separate xattr.
347 /** Bitfield for supported data in this structure. For future use. */
350 /** Incompat feature list. The supported feature mask is available in
351 * SOM_INCOMPAT_SUPP */
354 /** IO Epoch SOM attributes belongs to */
356 /** total file size in objects */
358 /** total fs blocks in objects */
360 /** mds mount id the size is valid for */
363 extern void lustre_som_swab(struct som_attrs
*attrs
);
365 #define SOM_INCOMPAT_SUPP 0x0
368 * HSM on-disk attributes stored in a separate xattr.
371 /** Bitfield for supported data in this structure. For future use. */
374 /** HSM flags, see hsm_flags enum below */
376 /** backend archive id associated with the file */
378 /** version associated with the last archiving, if any */
381 extern void lustre_hsm_swab(struct hsm_attrs
*attrs
);
387 /** LASTID file has zero OID */
388 LUSTRE_FID_LASTID_OID
= 0UL,
389 /** initial fid id value */
390 LUSTRE_FID_INIT_OID
= 1UL
393 /** returns fid object sequence */
394 static inline __u64
fid_seq(const struct lu_fid
*fid
)
399 /** returns fid object id */
400 static inline __u32
fid_oid(const struct lu_fid
*fid
)
405 /** returns fid object version */
406 static inline __u32
fid_ver(const struct lu_fid
*fid
)
411 static inline void fid_zero(struct lu_fid
*fid
)
413 memset(fid
, 0, sizeof(*fid
));
416 static inline obd_id
fid_ver_oid(const struct lu_fid
*fid
)
418 return ((__u64
)fid_ver(fid
) << 32 | fid_oid(fid
));
422 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
423 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
424 * used for other purposes and not risk collisions with existing inodes.
426 * Different FID Format
427 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
430 FID_SEQ_OST_MDT0
= 0,
431 FID_SEQ_LLOG
= 1, /* unnamed llogs */
433 FID_SEQ_OST_MDT1
= 3,
434 FID_SEQ_OST_MAX
= 9, /* Max MDT count before OST_on_FID */
435 FID_SEQ_LLOG_NAME
= 10, /* named llogs */
438 FID_SEQ_IGIF_MAX
= 0x0ffffffffULL
,
439 FID_SEQ_IDIF
= 0x100000000ULL
,
440 FID_SEQ_IDIF_MAX
= 0x1ffffffffULL
,
441 /* Normal FID sequence starts from this value, i.e. 1<<33 */
442 FID_SEQ_START
= 0x200000000ULL
,
443 /* sequence for local pre-defined FIDs listed in local_oid */
444 FID_SEQ_LOCAL_FILE
= 0x200000001ULL
,
445 FID_SEQ_DOT_LUSTRE
= 0x200000002ULL
,
446 /* sequence is used for local named objects FIDs generated
447 * by local_object_storage library */
448 FID_SEQ_LOCAL_NAME
= 0x200000003ULL
,
449 /* Because current FLD will only cache the fid sequence, instead
450 * of oid on the client side, if the FID needs to be exposed to
451 * clients sides, it needs to make sure all of fids under one
452 * sequence will be located in one MDT. */
453 FID_SEQ_SPECIAL
= 0x200000004ULL
,
454 FID_SEQ_QUOTA
= 0x200000005ULL
,
455 FID_SEQ_QUOTA_GLB
= 0x200000006ULL
,
456 FID_SEQ_ROOT
= 0x200000007ULL
, /* Located on MDT0 */
457 FID_SEQ_NORMAL
= 0x200000400ULL
,
458 FID_SEQ_LOV_DEFAULT
= 0xffffffffffffffffULL
461 #define OBIF_OID_MAX_BITS 32
462 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
463 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
464 #define IDIF_OID_MAX_BITS 48
465 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
466 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
468 /** OID for FID_SEQ_SPECIAL */
470 /* Big Filesystem Lock to serialize rename operations */
471 FID_OID_SPECIAL_BFL
= 1UL,
474 /** OID for FID_SEQ_DOT_LUSTRE */
475 enum dot_lustre_oid
{
476 FID_OID_DOT_LUSTRE
= 1UL,
477 FID_OID_DOT_LUSTRE_OBF
= 2UL,
480 static inline int fid_seq_is_mdt0(obd_seq seq
)
482 return (seq
== FID_SEQ_OST_MDT0
);
485 static inline int fid_seq_is_mdt(const __u64 seq
)
487 return seq
== FID_SEQ_OST_MDT0
|| seq
>= FID_SEQ_NORMAL
;
490 static inline int fid_seq_is_echo(obd_seq seq
)
492 return (seq
== FID_SEQ_ECHO
);
495 static inline int fid_is_echo(const struct lu_fid
*fid
)
497 return fid_seq_is_echo(fid_seq(fid
));
500 static inline int fid_seq_is_llog(obd_seq seq
)
502 return (seq
== FID_SEQ_LLOG
);
505 static inline int fid_is_llog(const struct lu_fid
*fid
)
507 /* file with OID == 0 is not llog but contains last oid */
508 return fid_seq_is_llog(fid_seq(fid
)) && fid_oid(fid
) > 0;
511 static inline int fid_seq_is_rsvd(const __u64 seq
)
513 return (seq
> FID_SEQ_OST_MDT0
&& seq
<= FID_SEQ_RSVD
);
516 static inline int fid_seq_is_special(const __u64 seq
)
518 return seq
== FID_SEQ_SPECIAL
;
521 static inline int fid_seq_is_local_file(const __u64 seq
)
523 return seq
== FID_SEQ_LOCAL_FILE
||
524 seq
== FID_SEQ_LOCAL_NAME
;
527 static inline int fid_seq_is_root(const __u64 seq
)
529 return seq
== FID_SEQ_ROOT
;
532 static inline int fid_seq_is_dot(const __u64 seq
)
534 return seq
== FID_SEQ_DOT_LUSTRE
;
537 static inline int fid_seq_is_default(const __u64 seq
)
539 return seq
== FID_SEQ_LOV_DEFAULT
;
542 static inline int fid_is_mdt0(const struct lu_fid
*fid
)
544 return fid_seq_is_mdt0(fid_seq(fid
));
547 static inline void lu_root_fid(struct lu_fid
*fid
)
549 fid
->f_seq
= FID_SEQ_ROOT
;
555 * Check if a fid is igif or not.
556 * \param fid the fid to be tested.
557 * \return true if the fid is a igif; otherwise false.
559 static inline int fid_seq_is_igif(const __u64 seq
)
561 return seq
>= FID_SEQ_IGIF
&& seq
<= FID_SEQ_IGIF_MAX
;
564 static inline int fid_is_igif(const struct lu_fid
*fid
)
566 return fid_seq_is_igif(fid_seq(fid
));
570 * Check if a fid is idif or not.
571 * \param fid the fid to be tested.
572 * \return true if the fid is a idif; otherwise false.
574 static inline int fid_seq_is_idif(const __u64 seq
)
576 return seq
>= FID_SEQ_IDIF
&& seq
<= FID_SEQ_IDIF_MAX
;
579 static inline int fid_is_idif(const struct lu_fid
*fid
)
581 return fid_seq_is_idif(fid_seq(fid
));
584 static inline int fid_is_local_file(const struct lu_fid
*fid
)
586 return fid_seq_is_local_file(fid_seq(fid
));
589 static inline int fid_seq_is_norm(const __u64 seq
)
591 return (seq
>= FID_SEQ_NORMAL
);
594 static inline int fid_is_norm(const struct lu_fid
*fid
)
596 return fid_seq_is_norm(fid_seq(fid
));
599 /* convert an OST objid into an IDIF FID SEQ number */
600 static inline obd_seq
fid_idif_seq(obd_id id
, __u32 ost_idx
)
602 return FID_SEQ_IDIF
| (ost_idx
<< 16) | ((id
>> 32) & 0xffff);
605 /* convert a packed IDIF FID into an OST objid */
606 static inline obd_id
fid_idif_id(obd_seq seq
, __u32 oid
, __u32 ver
)
608 return ((__u64
)ver
<< 48) | ((seq
& 0xffff) << 32) | oid
;
611 /* extract ost index from IDIF FID */
612 static inline __u32
fid_idif_ost_idx(const struct lu_fid
*fid
)
614 return (fid_seq(fid
) >> 16) & 0xffff;
617 /* extract OST sequence (group) from a wire ost_id (id/seq) pair */
618 static inline obd_seq
ostid_seq(const struct ost_id
*ostid
)
620 if (fid_seq_is_mdt0(ostid
->oi
.oi_seq
))
621 return FID_SEQ_OST_MDT0
;
623 if (fid_seq_is_default(ostid
->oi
.oi_seq
))
624 return FID_SEQ_LOV_DEFAULT
;
626 if (fid_is_idif(&ostid
->oi_fid
))
627 return FID_SEQ_OST_MDT0
;
629 return fid_seq(&ostid
->oi_fid
);
632 /* extract OST objid from a wire ost_id (id/seq) pair */
633 static inline obd_id
ostid_id(const struct ost_id
*ostid
)
635 if (fid_seq_is_mdt0(ostid_seq(ostid
)))
636 return ostid
->oi
.oi_id
& IDIF_OID_MASK
;
638 if (fid_is_idif(&ostid
->oi_fid
))
639 return fid_idif_id(fid_seq(&ostid
->oi_fid
),
640 fid_oid(&ostid
->oi_fid
), 0);
642 return fid_oid(&ostid
->oi_fid
);
645 static inline void ostid_set_seq(struct ost_id
*oi
, __u64 seq
)
647 if (fid_seq_is_mdt0(seq
) || fid_seq_is_default(seq
)) {
650 oi
->oi_fid
.f_seq
= seq
;
651 /* Note: if f_oid + f_ver is zero, we need init it
652 * to be 1, otherwise, ostid_seq will treat this
653 * as old ostid (oi_seq == 0) */
654 if (oi
->oi_fid
.f_oid
== 0 && oi
->oi_fid
.f_ver
== 0)
655 oi
->oi_fid
.f_oid
= LUSTRE_FID_INIT_OID
;
659 static inline void ostid_set_seq_mdt0(struct ost_id
*oi
)
661 ostid_set_seq(oi
, FID_SEQ_OST_MDT0
);
664 static inline void ostid_set_seq_echo(struct ost_id
*oi
)
666 ostid_set_seq(oi
, FID_SEQ_ECHO
);
669 static inline void ostid_set_seq_llog(struct ost_id
*oi
)
671 ostid_set_seq(oi
, FID_SEQ_LLOG
);
675 * Note: we need check oi_seq to decide where to set oi_id,
676 * so oi_seq should always be set ahead of oi_id.
678 static inline void ostid_set_id(struct ost_id
*oi
, __u64 oid
)
680 if (fid_seq_is_mdt0(ostid_seq(oi
))) {
681 if (oid
>= IDIF_MAX_OID
) {
682 CERROR("Bad %llu to set "DOSTID
"\n",
688 if (oid
> OBIF_MAX_OID
) {
689 CERROR("Bad %llu to set "DOSTID
"\n",
693 oi
->oi_fid
.f_oid
= oid
;
697 static inline void ostid_inc_id(struct ost_id
*oi
)
699 if (fid_seq_is_mdt0(ostid_seq(oi
))) {
700 if (unlikely(ostid_id(oi
) + 1 > IDIF_MAX_OID
)) {
701 CERROR("Bad inc "DOSTID
"\n", POSTID(oi
));
710 static inline void ostid_dec_id(struct ost_id
*oi
)
712 if (fid_seq_is_mdt0(ostid_seq(oi
)))
719 * Unpack an OST object id/seq (group) into a FID. This is needed for
720 * converting all obdo, lmm, lsm, etc. 64-bit id/seq pairs into proper
721 * FIDs. Note that if an id/seq is already in FID/IDIF format it will
722 * be passed through unchanged. Only legacy OST objects in "group 0"
723 * will be mapped into the IDIF namespace so that they can fit into the
724 * struct lu_fid fields without loss. For reference see:
725 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs
727 static inline int ostid_to_fid(struct lu_fid
*fid
, struct ost_id
*ostid
,
730 if (ost_idx
> 0xffff) {
731 CERROR("bad ost_idx, "DOSTID
" ost_idx:%u\n", POSTID(ostid
),
736 if (fid_seq_is_mdt0(ostid_seq(ostid
))) {
737 /* This is a "legacy" (old 1.x/2.early) OST object in "group 0"
738 * that we map into the IDIF namespace. It allows up to 2^48
739 * objects per OST, as this is the object namespace that has
740 * been in production for years. This can handle create rates
741 * of 1M objects/s/OST for 9 years, or combinations thereof. */
742 if (ostid_id(ostid
) >= IDIF_MAX_OID
) {
743 CERROR("bad MDT0 id, "DOSTID
" ost_idx:%u\n",
744 POSTID(ostid
), ost_idx
);
747 fid
->f_seq
= fid_idif_seq(ostid_id(ostid
), ost_idx
);
748 /* truncate to 32 bits by assignment */
749 fid
->f_oid
= ostid_id(ostid
);
750 /* in theory, not currently used */
751 fid
->f_ver
= ostid_id(ostid
) >> 48;
752 } else /* if (fid_seq_is_idif(seq) || fid_seq_is_norm(seq)) */ {
753 /* This is either an IDIF object, which identifies objects across
754 * all OSTs, or a regular FID. The IDIF namespace maps legacy
755 * OST objects into the FID namespace. In both cases, we just
756 * pass the FID through, no conversion needed. */
757 if (ostid
->oi_fid
.f_ver
!= 0) {
758 CERROR("bad MDT0 id, "DOSTID
" ost_idx:%u\n",
759 POSTID(ostid
), ost_idx
);
762 *fid
= ostid
->oi_fid
;
768 /* pack any OST FID into an ostid (id/seq) for the wire/disk */
769 static inline int fid_to_ostid(const struct lu_fid
*fid
, struct ost_id
*ostid
)
771 if (unlikely(fid_seq_is_igif(fid
->f_seq
))) {
772 CERROR("bad IGIF, "DFID
"\n", PFID(fid
));
776 if (fid_is_idif(fid
)) {
777 ostid_set_seq_mdt0(ostid
);
778 ostid_set_id(ostid
, fid_idif_id(fid_seq(fid
), fid_oid(fid
),
781 ostid
->oi_fid
= *fid
;
787 /* Check whether the fid is for LAST_ID */
788 static inline int fid_is_last_id(const struct lu_fid
*fid
)
790 return (fid_oid(fid
) == 0);
794 * Get inode number from a igif.
795 * \param fid a igif to get inode number from.
796 * \return inode number for the igif.
798 static inline ino_t
lu_igif_ino(const struct lu_fid
*fid
)
803 extern void lustre_swab_ost_id(struct ost_id
*oid
);
806 * Get inode generation from a igif.
807 * \param fid a igif to get inode generation from.
808 * \return inode generation for the igif.
810 static inline __u32
lu_igif_gen(const struct lu_fid
*fid
)
816 * Build igif from the inode number/generation.
818 static inline void lu_igif_build(struct lu_fid
*fid
, __u32 ino
, __u32 gen
)
826 * Fids are transmitted across network (in the sender byte-ordering),
827 * and stored on disk in big-endian order.
829 static inline void fid_cpu_to_le(struct lu_fid
*dst
, const struct lu_fid
*src
)
831 dst
->f_seq
= cpu_to_le64(fid_seq(src
));
832 dst
->f_oid
= cpu_to_le32(fid_oid(src
));
833 dst
->f_ver
= cpu_to_le32(fid_ver(src
));
836 static inline void fid_le_to_cpu(struct lu_fid
*dst
, const struct lu_fid
*src
)
838 dst
->f_seq
= le64_to_cpu(fid_seq(src
));
839 dst
->f_oid
= le32_to_cpu(fid_oid(src
));
840 dst
->f_ver
= le32_to_cpu(fid_ver(src
));
843 static inline void fid_cpu_to_be(struct lu_fid
*dst
, const struct lu_fid
*src
)
845 dst
->f_seq
= cpu_to_be64(fid_seq(src
));
846 dst
->f_oid
= cpu_to_be32(fid_oid(src
));
847 dst
->f_ver
= cpu_to_be32(fid_ver(src
));
850 static inline void fid_be_to_cpu(struct lu_fid
*dst
, const struct lu_fid
*src
)
852 dst
->f_seq
= be64_to_cpu(fid_seq(src
));
853 dst
->f_oid
= be32_to_cpu(fid_oid(src
));
854 dst
->f_ver
= be32_to_cpu(fid_ver(src
));
857 static inline int fid_is_sane(const struct lu_fid
*fid
)
859 return fid
!= NULL
&&
860 ((fid_seq(fid
) >= FID_SEQ_START
&& fid_ver(fid
) == 0) ||
861 fid_is_igif(fid
) || fid_is_idif(fid
) ||
862 fid_seq_is_rsvd(fid_seq(fid
)));
865 static inline int fid_is_zero(const struct lu_fid
*fid
)
867 return fid_seq(fid
) == 0 && fid_oid(fid
) == 0;
870 extern void lustre_swab_lu_fid(struct lu_fid
*fid
);
871 extern void lustre_swab_lu_seq_range(struct lu_seq_range
*range
);
873 static inline int lu_fid_eq(const struct lu_fid
*f0
, const struct lu_fid
*f1
)
875 return memcmp(f0
, f1
, sizeof(*f0
)) == 0;
878 #define __diff_normalize(val0, val1) \
880 typeof(val0) __val0 = (val0); \
881 typeof(val1) __val1 = (val1); \
883 (__val0 == __val1 ? 0 : __val0 > __val1 ? +1 : -1); \
886 static inline int lu_fid_cmp(const struct lu_fid
*f0
,
887 const struct lu_fid
*f1
)
890 __diff_normalize(fid_seq(f0
), fid_seq(f1
)) ?:
891 __diff_normalize(fid_oid(f0
), fid_oid(f1
)) ?:
892 __diff_normalize(fid_ver(f0
), fid_ver(f1
));
895 static inline void ostid_cpu_to_le(const struct ost_id
*src_oi
,
896 struct ost_id
*dst_oi
)
898 if (fid_seq_is_mdt0(ostid_seq(src_oi
))) {
899 dst_oi
->oi
.oi_id
= cpu_to_le64(src_oi
->oi
.oi_id
);
900 dst_oi
->oi
.oi_seq
= cpu_to_le64(src_oi
->oi
.oi_seq
);
902 fid_cpu_to_le(&dst_oi
->oi_fid
, &src_oi
->oi_fid
);
906 static inline void ostid_le_to_cpu(const struct ost_id
*src_oi
,
907 struct ost_id
*dst_oi
)
909 if (fid_seq_is_mdt0(ostid_seq(src_oi
))) {
910 dst_oi
->oi
.oi_id
= le64_to_cpu(src_oi
->oi
.oi_id
);
911 dst_oi
->oi
.oi_seq
= le64_to_cpu(src_oi
->oi
.oi_seq
);
913 fid_le_to_cpu(&dst_oi
->oi_fid
, &src_oi
->oi_fid
);
919 /** \defgroup lu_dir lu_dir
923 * Enumeration of possible directory entry attributes.
925 * Attributes follow directory entry header in the order they appear in this
928 enum lu_dirent_attrs
{
931 LUDA_64BITHASH
= 0x0004,
933 /* The following attrs are used for MDT internal only,
934 * not visible to client */
936 /* Verify the dirent consistency */
937 LUDA_VERIFY
= 0x8000,
938 /* Only check but not repair the dirent inconsistency */
939 LUDA_VERIFY_DRYRUN
= 0x4000,
940 /* The dirent has been repaired, or to be repaired (dryrun). */
941 LUDA_REPAIR
= 0x2000,
942 /* The system is upgraded, has beed or to be repaired (dryrun). */
943 LUDA_UPGRADE
= 0x1000,
944 /* Ignore this record, go to next directly. */
945 LUDA_IGNORE
= 0x0800,
948 #define LU_DIRENT_ATTRS_MASK 0xf800
951 * Layout of readdir pages, as transmitted on wire.
954 /** valid if LUDA_FID is set. */
955 struct lu_fid lde_fid
;
956 /** a unique entry identifier: a hash or an offset. */
958 /** total record length, including all attributes. */
962 /** optional variable size attributes following this entry.
963 * taken from enum lu_dirent_attrs.
966 /** name is followed by the attributes indicated in ->ldp_attrs, in
967 * their natural order. After the last attribute, padding bytes are
968 * added to make ->lde_reclen a multiple of 8.
974 * Definitions of optional directory entry attributes formats.
976 * Individual attributes do not have their length encoded in a generic way. It
977 * is assumed that consumer of an attribute knows its format. This means that
978 * it is impossible to skip over an unknown attribute, except by skipping over all
979 * remaining attributes (by using ->lde_reclen), which is not too
980 * constraining, because new server versions will append new attributes at
981 * the end of an entry.
985 * Fid directory attribute: a fid of an object referenced by the entry. This
986 * will be almost always requested by the client and supplied by the server.
988 * Aligned to 8 bytes.
990 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
995 * Aligned to 2 bytes.
1006 #define IFTODT(type) (((type) & S_IFMT) >> IFSHIFT)
1009 #define DTTOIF(dirtype) ((dirtype) << IFSHIFT)
1014 __u64 ldp_hash_start
;
1018 struct lu_dirent ldp_entries
[0];
1021 enum lu_dirpage_flags
{
1023 * dirpage contains no entry.
1027 * last entry's lde_hash equals ldp_hash_end.
1029 LDF_COLLIDE
= 1 << 1
1032 static inline struct lu_dirent
*lu_dirent_start(struct lu_dirpage
*dp
)
1034 if (le32_to_cpu(dp
->ldp_flags
) & LDF_EMPTY
)
1037 return dp
->ldp_entries
;
1040 static inline struct lu_dirent
*lu_dirent_next(struct lu_dirent
*ent
)
1042 struct lu_dirent
*next
;
1044 if (le16_to_cpu(ent
->lde_reclen
) != 0)
1045 next
= ((void *)ent
) + le16_to_cpu(ent
->lde_reclen
);
1052 static inline int lu_dirent_calc_size(int namelen
, __u16 attr
)
1056 if (attr
& LUDA_TYPE
) {
1057 const unsigned align
= sizeof(struct luda_type
) - 1;
1058 size
= (sizeof(struct lu_dirent
) + namelen
+ align
) & ~align
;
1059 size
+= sizeof(struct luda_type
);
1061 size
= sizeof(struct lu_dirent
) + namelen
;
1063 return (size
+ 7) & ~7;
1066 static inline int lu_dirent_size(struct lu_dirent
*ent
)
1068 if (le16_to_cpu(ent
->lde_reclen
) == 0) {
1069 return lu_dirent_calc_size(le16_to_cpu(ent
->lde_namelen
),
1070 le32_to_cpu(ent
->lde_attrs
));
1072 return le16_to_cpu(ent
->lde_reclen
);
1075 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
1078 * MDS_READPAGE page size
1080 * This is the directory page size packed in MDS_READPAGE RPC.
1081 * It's different than PAGE_CACHE_SIZE because the client needs to
1082 * access the struct lu_dirpage header packed at the beginning of
1083 * the "page" and without this there isn't any way to know find the
1084 * lu_dirpage header is if client and server PAGE_CACHE_SIZE differ.
1086 #define LU_PAGE_SHIFT 12
1087 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
1088 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
1090 #define LU_PAGE_COUNT (1 << (PAGE_CACHE_SHIFT - LU_PAGE_SHIFT))
1094 struct lustre_handle
{
1097 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
1099 static inline int lustre_handle_is_used(struct lustre_handle
*lh
)
1101 return lh
->cookie
!= 0ull;
1104 static inline int lustre_handle_equal(const struct lustre_handle
*lh1
,
1105 const struct lustre_handle
*lh2
)
1107 return lh1
->cookie
== lh2
->cookie
;
1110 static inline void lustre_handle_copy(struct lustre_handle
*tgt
,
1111 struct lustre_handle
*src
)
1113 tgt
->cookie
= src
->cookie
;
1116 /* flags for lm_flags */
1117 #define MSGHDR_AT_SUPPORT 0x1
1118 #define MSGHDR_CKSUM_INCOMPAT18 0x2
1120 #define lustre_msg lustre_msg_v2
1121 /* we depend on this structure to be 8-byte aligned */
1122 /* this type is only endian-adjusted in lustre_unpack_msg() */
1123 struct lustre_msg_v2
{
1132 __u32 lm_buflens
[0];
1135 /* without gss, ptlrpc_body is put at the first buffer. */
1136 #define PTLRPC_NUM_VERSIONS 4
1137 #define JOBSTATS_JOBID_SIZE 32 /* 32 bytes string */
1138 struct ptlrpc_body_v3
{
1139 struct lustre_handle pb_handle
;
1146 __u64 pb_last_committed
;
1151 __u32 pb_timeout
; /* for req, the deadline, for rep, the service est */
1152 __u32 pb_service_time
; /* for rep, actual service time */
1155 /* VBR: pre-versions */
1156 __u64 pb_pre_versions
[PTLRPC_NUM_VERSIONS
];
1157 /* padding for future needs */
1158 __u64 pb_padding
[4];
1159 char pb_jobid
[JOBSTATS_JOBID_SIZE
];
1161 #define ptlrpc_body ptlrpc_body_v3
1163 struct ptlrpc_body_v2
{
1164 struct lustre_handle pb_handle
;
1171 __u64 pb_last_committed
;
1176 __u32 pb_timeout
; /* for req, the deadline, for rep, the service est */
1177 __u32 pb_service_time
; /* for rep, actual service time, also used for
1178 net_latency of req */
1181 /* VBR: pre-versions */
1182 __u64 pb_pre_versions
[PTLRPC_NUM_VERSIONS
];
1183 /* padding for future needs */
1184 __u64 pb_padding
[4];
1187 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body
*pb
);
1189 /* message body offset for lustre_msg_v2 */
1190 /* ptlrpc body offset in all request/reply messages */
1191 #define MSG_PTLRPC_BODY_OFF 0
1193 /* normal request/reply message record offset */
1194 #define REQ_REC_OFF 1
1195 #define REPLY_REC_OFF 1
1197 /* ldlm request message body offset */
1198 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
1199 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
1201 /* ldlm intent lock message body offset */
1202 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
1203 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
1205 /* ldlm reply message body offset */
1206 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
1207 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
1209 /** only use in req->rq_{req,rep}_swab_mask */
1210 #define MSG_PTLRPC_HEADER_OFF 31
1212 /* Flags that are operation-specific go in the top 16 bits. */
1213 #define MSG_OP_FLAG_MASK 0xffff0000
1214 #define MSG_OP_FLAG_SHIFT 16
1216 /* Flags that apply to all requests are in the bottom 16 bits */
1217 #define MSG_GEN_FLAG_MASK 0x0000ffff
1218 #define MSG_LAST_REPLAY 0x0001
1219 #define MSG_RESENT 0x0002
1220 #define MSG_REPLAY 0x0004
1221 /* #define MSG_AT_SUPPORT 0x0008
1222 * This was used in early prototypes of adaptive timeouts, and while there
1223 * shouldn't be any users of that code there also isn't a need for using this
1224 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
1225 #define MSG_DELAY_REPLAY 0x0010
1226 #define MSG_VERSION_REPLAY 0x0020
1227 #define MSG_REQ_REPLAY_DONE 0x0040
1228 #define MSG_LOCK_REPLAY_DONE 0x0080
1231 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
1234 #define MSG_CONNECT_RECOVERING 0x00000001
1235 #define MSG_CONNECT_RECONNECT 0x00000002
1236 #define MSG_CONNECT_REPLAYABLE 0x00000004
1237 //#define MSG_CONNECT_PEER 0x8
1238 #define MSG_CONNECT_LIBCLIENT 0x00000010
1239 #define MSG_CONNECT_INITIAL 0x00000020
1240 #define MSG_CONNECT_ASYNC 0x00000040
1241 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
1242 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
1245 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
1246 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
1247 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
1248 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
1249 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
1250 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
1251 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
1252 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
1253 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
1254 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
1255 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
1256 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
1257 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
1258 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
1259 *We do not support JOIN FILE
1260 *anymore, reserve this flags
1261 *just for preventing such bit
1263 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
1264 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
1265 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /*Remote client */
1266 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /*Remote client by force */
1267 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
1268 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
1269 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
1270 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
1271 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
1272 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
1273 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
1274 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
1275 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
1276 #define OBD_CONNECT_REAL 0x8000000ULL /*real connection */
1277 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
1278 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
1279 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
1280 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
1281 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
1282 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
1283 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
1284 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
1285 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
1286 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
1287 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
1289 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
1290 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
1291 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
1292 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
1293 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
1294 * RPC error properly */
1295 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
1296 * finer space reservation */
1297 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
1298 * policy and 2.x server */
1299 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
1300 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
1301 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
1302 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
1303 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
1304 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* flock deadlock detection */
1305 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/*create stripe disposition*/
1308 * Please DO NOT add flag values here before first ensuring that this same
1309 * flag value is not in use on some other branch. Please clear any such
1310 * changes with senior engineers before starting to use a new flag. Then,
1311 * submit a small patch against EVERY branch that ONLY adds the new flag,
1312 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
1313 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
1314 * can be approved and landed easily to reserve the flag for future use. */
1316 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
1317 * connection. It is a temporary bug fix for Imperative Recovery interop
1318 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
1319 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
1320 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
1322 #define OCD_HAS_FLAG(ocd, flg) \
1323 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
1326 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
1328 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
1329 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
1330 OBD_CONNECT_IBITS | \
1331 OBD_CONNECT_NODEVOH | OBD_CONNECT_ATTRFID | \
1332 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1333 OBD_CONNECT_RMT_CLIENT | \
1334 OBD_CONNECT_RMT_CLIENT_FORCE | \
1335 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_MDS_CAPA | \
1336 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_MDS_MDS | \
1337 OBD_CONNECT_FID | LRU_RESIZE_CONNECT_FLAG | \
1338 OBD_CONNECT_VBR | OBD_CONNECT_LOV_V3 | \
1339 OBD_CONNECT_SOM | OBD_CONNECT_FULL20 | \
1340 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
1341 OBD_CONNECT_EINPROGRESS | \
1342 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
1343 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
1344 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
1345 OBD_CONNECT_FLOCK_DEAD | \
1346 OBD_CONNECT_DISP_STRIPE)
1348 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
1349 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
1350 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
1351 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_OSS_CAPA | \
1352 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1353 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_CKSUM | \
1354 OBD_CONNECT_RMT_CLIENT | \
1355 OBD_CONNECT_RMT_CLIENT_FORCE | OBD_CONNECT_VBR | \
1356 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
1357 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 | \
1358 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
1359 OBD_CONNECT_MAX_EASIZE | \
1360 OBD_CONNECT_EINPROGRESS | \
1361 OBD_CONNECT_JOBSTATS | \
1362 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
1363 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
1364 OBD_CONNECT_PINGLESS)
1365 #define ECHO_CONNECT_SUPPORTED (0)
1366 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
1367 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
1368 OBD_CONNECT_MNE_SWAB | OBD_CONNECT_PINGLESS)
1370 /* Features required for this version of the client to work with server */
1371 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
1374 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
1375 ((patch)<<8) + (fix))
1376 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
1377 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
1378 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
1379 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
1381 /* This structure is used for both request and reply.
1383 * If we eventually have separate connect data for different types, which we
1384 * almost certainly will, then perhaps we stick a union in here. */
1385 struct obd_connect_data_v1
{
1386 __u64 ocd_connect_flags
; /* OBD_CONNECT_* per above */
1387 __u32 ocd_version
; /* lustre release version number */
1388 __u32 ocd_grant
; /* initial cache grant amount (bytes) */
1389 __u32 ocd_index
; /* LOV index to connect to */
1390 __u32 ocd_brw_size
; /* Maximum BRW size in bytes, must be 2^n */
1391 __u64 ocd_ibits_known
; /* inode bits this client understands */
1392 __u8 ocd_blocksize
; /* log2 of the backend filesystem blocksize */
1393 __u8 ocd_inodespace
; /* log2 of the per-inode space consumption */
1394 __u16 ocd_grant_extent
; /* per-extent grant overhead, in 1K blocks */
1395 __u32 ocd_unused
; /* also fix lustre_swab_connect */
1396 __u64 ocd_transno
; /* first transno from client to be replayed */
1397 __u32 ocd_group
; /* MDS group on OST */
1398 __u32 ocd_cksum_types
; /* supported checksum algorithms */
1399 __u32 ocd_max_easize
; /* How big LOV EA can be on MDS */
1400 __u32 ocd_instance
; /* also fix lustre_swab_connect */
1401 __u64 ocd_maxbytes
; /* Maximum stripe size in bytes */
1404 struct obd_connect_data
{
1405 __u64 ocd_connect_flags
; /* OBD_CONNECT_* per above */
1406 __u32 ocd_version
; /* lustre release version number */
1407 __u32 ocd_grant
; /* initial cache grant amount (bytes) */
1408 __u32 ocd_index
; /* LOV index to connect to */
1409 __u32 ocd_brw_size
; /* Maximum BRW size in bytes */
1410 __u64 ocd_ibits_known
; /* inode bits this client understands */
1411 __u8 ocd_blocksize
; /* log2 of the backend filesystem blocksize */
1412 __u8 ocd_inodespace
; /* log2 of the per-inode space consumption */
1413 __u16 ocd_grant_extent
; /* per-extent grant overhead, in 1K blocks */
1414 __u32 ocd_unused
; /* also fix lustre_swab_connect */
1415 __u64 ocd_transno
; /* first transno from client to be replayed */
1416 __u32 ocd_group
; /* MDS group on OST */
1417 __u32 ocd_cksum_types
; /* supported checksum algorithms */
1418 __u32 ocd_max_easize
; /* How big LOV EA can be on MDS */
1419 __u32 ocd_instance
; /* instance # of this target */
1420 __u64 ocd_maxbytes
; /* Maximum stripe size in bytes */
1421 /* Fields after ocd_maxbytes are only accessible by the receiver
1422 * if the corresponding flag in ocd_connect_flags is set. Accessing
1423 * any field after ocd_maxbytes on the receiver without a valid flag
1424 * may result in out-of-bound memory access and kernel oops. */
1425 __u64 padding1
; /* added 2.1.0. also fix lustre_swab_connect */
1426 __u64 padding2
; /* added 2.1.0. also fix lustre_swab_connect */
1427 __u64 padding3
; /* added 2.1.0. also fix lustre_swab_connect */
1428 __u64 padding4
; /* added 2.1.0. also fix lustre_swab_connect */
1429 __u64 padding5
; /* added 2.1.0. also fix lustre_swab_connect */
1430 __u64 padding6
; /* added 2.1.0. also fix lustre_swab_connect */
1431 __u64 padding7
; /* added 2.1.0. also fix lustre_swab_connect */
1432 __u64 padding8
; /* added 2.1.0. also fix lustre_swab_connect */
1433 __u64 padding9
; /* added 2.1.0. also fix lustre_swab_connect */
1434 __u64 paddingA
; /* added 2.1.0. also fix lustre_swab_connect */
1435 __u64 paddingB
; /* added 2.1.0. also fix lustre_swab_connect */
1436 __u64 paddingC
; /* added 2.1.0. also fix lustre_swab_connect */
1437 __u64 paddingD
; /* added 2.1.0. also fix lustre_swab_connect */
1438 __u64 paddingE
; /* added 2.1.0. also fix lustre_swab_connect */
1439 __u64 paddingF
; /* added 2.1.0. also fix lustre_swab_connect */
1442 * Please DO NOT use any fields here before first ensuring that this same
1443 * field is not in use on some other branch. Please clear any such changes
1444 * with senior engineers before starting to use a new field. Then, submit
1445 * a small patch against EVERY branch that ONLY adds the new field along with
1446 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
1447 * reserve the flag for future use. */
1450 extern void lustre_swab_connect(struct obd_connect_data
*ocd
);
1453 * Supported checksum algorithms. Up to 32 checksum types are supported.
1454 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1455 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
1456 * algorithm and also the OBD_FL_CKSUM* flags.
1459 OBD_CKSUM_CRC32
= 0x00000001,
1460 OBD_CKSUM_ADLER
= 0x00000002,
1461 OBD_CKSUM_CRC32C
= 0x00000004,
1465 * OST requests: OBDO & OBD request records
1470 OST_REPLY
= 0, /* reply ? */
1486 OST_QUOTACHECK
= 18,
1488 OST_QUOTA_ADJUST_QUNIT
= 20, /* not used since 2.4 */
1491 #define OST_FIRST_OPC OST_REPLY
1494 OBD_FL_INLINEDATA
= 0x00000001,
1495 OBD_FL_OBDMDEXISTS
= 0x00000002,
1496 OBD_FL_DELORPHAN
= 0x00000004, /* if set in o_flags delete orphans */
1497 OBD_FL_NORPC
= 0x00000008, /* set in o_flags do in OSC not OST */
1498 OBD_FL_IDONLY
= 0x00000010, /* set in o_flags only adjust obj id*/
1499 OBD_FL_RECREATE_OBJS
= 0x00000020, /* recreate missing obj */
1500 OBD_FL_DEBUG_CHECK
= 0x00000040, /* echo client/server debug check */
1501 OBD_FL_NO_USRQUOTA
= 0x00000100, /* the object's owner is over quota */
1502 OBD_FL_NO_GRPQUOTA
= 0x00000200, /* the object's group is over quota */
1503 OBD_FL_CREATE_CROW
= 0x00000400, /* object should be create on write */
1504 OBD_FL_SRVLOCK
= 0x00000800, /* delegate DLM locking to server */
1505 OBD_FL_CKSUM_CRC32
= 0x00001000, /* CRC32 checksum type */
1506 OBD_FL_CKSUM_ADLER
= 0x00002000, /* ADLER checksum type */
1507 OBD_FL_CKSUM_CRC32C
= 0x00004000, /* CRC32C checksum type */
1508 OBD_FL_CKSUM_RSVD2
= 0x00008000, /* for future cksum types */
1509 OBD_FL_CKSUM_RSVD3
= 0x00010000, /* for future cksum types */
1510 OBD_FL_SHRINK_GRANT
= 0x00020000, /* object shrink the grant */
1511 OBD_FL_MMAP
= 0x00040000, /* object is mmapped on the client.
1512 * XXX: obsoleted - reserved for old
1513 * clients prior than 2.2 */
1514 OBD_FL_RECOV_RESEND
= 0x00080000, /* recoverable resent */
1515 OBD_FL_NOSPC_BLK
= 0x00100000, /* no more block space on OST */
1517 /* Note that while these checksum values are currently separate bits,
1518 * in 2.x we can actually allow all values from 1-31 if we wanted. */
1519 OBD_FL_CKSUM_ALL
= OBD_FL_CKSUM_CRC32
| OBD_FL_CKSUM_ADLER
|
1520 OBD_FL_CKSUM_CRC32C
,
1522 /* mask for local-only flag, which won't be sent over network */
1523 OBD_FL_LOCAL_MASK
= 0xF0000000,
1526 #define LOV_MAGIC_V1 0x0BD10BD0
1527 #define LOV_MAGIC LOV_MAGIC_V1
1528 #define LOV_MAGIC_JOIN_V1 0x0BD20BD0
1529 #define LOV_MAGIC_V3 0x0BD30BD0
1532 * magic for fully defined striping
1533 * the idea is that we should have different magics for striping "hints"
1534 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1535 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1536 * we can't just change it w/o long way preparation, but we still need a
1537 * mechanism to allow LOD to differentiate hint versus ready striping.
1538 * so, at the moment we do a trick: MDT knows what to expect from request
1539 * depending on the case (replay uses ready striping, non-replay req uses
1540 * hints), so MDT replaces magic with appropriate one and now LOD can
1541 * easily understand what's inside -bzzz
1543 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
1544 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
1546 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
1547 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
1548 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
1549 #define LOV_PATTERN_CMOBD 0x200
1551 #define LOV_PATTERN_F_MASK 0xffff0000
1552 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
1554 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1555 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1557 #define lov_ost_data lov_ost_data_v1
1558 struct lov_ost_data_v1
{ /* per-stripe data structure (little-endian)*/
1559 struct ost_id l_ost_oi
; /* OST object ID */
1560 __u32 l_ost_gen
; /* generation of this l_ost_idx */
1561 __u32 l_ost_idx
; /* OST index in LOV (lov_tgt_desc->tgts) */
1564 #define lov_mds_md lov_mds_md_v1
1565 struct lov_mds_md_v1
{ /* LOV EA mds/wire data (little-endian) */
1566 __u32 lmm_magic
; /* magic number = LOV_MAGIC_V1 */
1567 __u32 lmm_pattern
; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1568 struct ost_id lmm_oi
; /* LOV object ID */
1569 __u32 lmm_stripe_size
; /* size of stripe in bytes */
1570 /* lmm_stripe_count used to be __u32 */
1571 __u16 lmm_stripe_count
; /* num stripes in use for this object */
1572 __u16 lmm_layout_gen
; /* layout generation number */
1573 struct lov_ost_data_v1 lmm_objects
[0]; /* per-stripe data */
1577 * Sigh, because pre-2.4 uses
1578 * struct lov_mds_md_v1 {
1580 * __u64 lmm_object_id;
1581 * __u64 lmm_object_seq;
1584 * to identify the LOV(MDT) object, and lmm_object_seq will
1585 * be normal_fid, which make it hard to combine these conversion
1586 * to ostid_to FID. so we will do lmm_oi/fid conversion separately
1588 * We can tell the lmm_oi by this way,
1589 * 1.8: lmm_object_id = {inode}, lmm_object_gr = 0
1590 * 2.1: lmm_object_id = {oid < 128k}, lmm_object_seq = FID_SEQ_NORMAL
1591 * 2.4: lmm_oi.f_seq = FID_SEQ_NORMAL, lmm_oi.f_oid = {oid < 128k},
1594 * But currently lmm_oi/lsm_oi does not have any "real" usages,
1595 * except for printing some information, and the user can always
1596 * get the real FID from LMA, besides this multiple case check might
1597 * make swab more complicate. So we will keep using id/seq for lmm_oi.
1600 static inline void fid_to_lmm_oi(const struct lu_fid
*fid
,
1603 oi
->oi
.oi_id
= fid_oid(fid
);
1604 oi
->oi
.oi_seq
= fid_seq(fid
);
1607 static inline void lmm_oi_set_seq(struct ost_id
*oi
, __u64 seq
)
1609 oi
->oi
.oi_seq
= seq
;
1612 static inline __u64
lmm_oi_id(struct ost_id
*oi
)
1614 return oi
->oi
.oi_id
;
1617 static inline __u64
lmm_oi_seq(struct ost_id
*oi
)
1619 return oi
->oi
.oi_seq
;
1622 static inline void lmm_oi_le_to_cpu(struct ost_id
*dst_oi
,
1623 struct ost_id
*src_oi
)
1625 dst_oi
->oi
.oi_id
= le64_to_cpu(src_oi
->oi
.oi_id
);
1626 dst_oi
->oi
.oi_seq
= le64_to_cpu(src_oi
->oi
.oi_seq
);
1629 static inline void lmm_oi_cpu_to_le(struct ost_id
*dst_oi
,
1630 struct ost_id
*src_oi
)
1632 dst_oi
->oi
.oi_id
= cpu_to_le64(src_oi
->oi
.oi_id
);
1633 dst_oi
->oi
.oi_seq
= cpu_to_le64(src_oi
->oi
.oi_seq
);
1636 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
1638 #define MAX_MD_SIZE \
1639 (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1640 #define MIN_MD_SIZE \
1641 (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1643 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1644 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1645 #define XATTR_USER_PREFIX "user."
1646 #define XATTR_TRUSTED_PREFIX "trusted."
1647 #define XATTR_SECURITY_PREFIX "security."
1648 #define XATTR_LUSTRE_PREFIX "lustre."
1650 #define XATTR_NAME_LOV "trusted.lov"
1651 #define XATTR_NAME_LMA "trusted.lma"
1652 #define XATTR_NAME_LMV "trusted.lmv"
1653 #define XATTR_NAME_LINK "trusted.link"
1654 #define XATTR_NAME_FID "trusted.fid"
1655 #define XATTR_NAME_VERSION "trusted.version"
1656 #define XATTR_NAME_SOM "trusted.som"
1657 #define XATTR_NAME_HSM "trusted.hsm"
1658 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_namespace"
1660 struct lov_mds_md_v3
{ /* LOV EA mds/wire data (little-endian) */
1661 __u32 lmm_magic
; /* magic number = LOV_MAGIC_V3 */
1662 __u32 lmm_pattern
; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1663 struct ost_id lmm_oi
; /* LOV object ID */
1664 __u32 lmm_stripe_size
; /* size of stripe in bytes */
1665 /* lmm_stripe_count used to be __u32 */
1666 __u16 lmm_stripe_count
; /* num stripes in use for this object */
1667 __u16 lmm_layout_gen
; /* layout generation number */
1668 char lmm_pool_name
[LOV_MAXPOOLNAME
]; /* must be 32bit aligned */
1669 struct lov_ost_data_v1 lmm_objects
[0]; /* per-stripe data */
1672 static inline __u32
lov_mds_md_size(__u16 stripes
, __u32 lmm_magic
)
1674 if (lmm_magic
== LOV_MAGIC_V3
)
1675 return sizeof(struct lov_mds_md_v3
) +
1676 stripes
* sizeof(struct lov_ost_data_v1
);
1678 return sizeof(struct lov_mds_md_v1
) +
1679 stripes
* sizeof(struct lov_ost_data_v1
);
1683 lov_mds_md_max_stripe_count(size_t buf_size
, __u32 lmm_magic
)
1685 switch (lmm_magic
) {
1686 case LOV_MAGIC_V1
: {
1687 struct lov_mds_md_v1 lmm
;
1689 if (buf_size
< sizeof(lmm
))
1692 return (buf_size
- sizeof(lmm
)) / sizeof(lmm
.lmm_objects
[0]);
1694 case LOV_MAGIC_V3
: {
1695 struct lov_mds_md_v3 lmm
;
1697 if (buf_size
< sizeof(lmm
))
1700 return (buf_size
- sizeof(lmm
)) / sizeof(lmm
.lmm_objects
[0]);
1707 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1708 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1709 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1710 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1711 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1712 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1713 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1714 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1715 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1716 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1717 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1718 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1719 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1720 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1721 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
1722 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1723 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1724 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1725 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1726 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1727 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1728 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1729 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
1730 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1731 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1732 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1733 /* ->mds if epoch opens or closes */
1734 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1735 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1736 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1737 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1738 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1740 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1741 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1742 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1743 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1745 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1746 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1747 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1748 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1749 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
1750 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1751 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1752 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1753 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1754 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1755 * under lock; for xattr
1756 * requests means the
1757 * client holds the lock */
1758 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1760 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
1761 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
1762 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
1763 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
1765 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1766 #define OBD_MD_FLRELEASED (0x0020000000000000ULL) /* file released */
1768 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1769 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1770 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1771 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1772 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1774 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1776 /* don't forget obdo_fid which is way down at the bottom so it can
1777 * come after the definition of llog_cookie */
1781 HSS_CLEARMASK
= 0x02,
1782 HSS_ARCHIVE_ID
= 0x04,
1785 struct hsm_state_set
{
1787 __u32 hss_archive_id
;
1789 __u64 hss_clearmask
;
1792 extern void lustre_swab_hsm_user_state(struct hsm_user_state
*hus
);
1793 extern void lustre_swab_hsm_state_set(struct hsm_state_set
*hss
);
1795 extern void lustre_swab_obd_statfs (struct obd_statfs
*os
);
1797 /* ost_body.data values for OST_BRW */
1799 #define OBD_BRW_READ 0x01
1800 #define OBD_BRW_WRITE 0x02
1801 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1802 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1803 * transfer and is not accounted in
1805 #define OBD_BRW_CHECK 0x10
1806 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1807 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1808 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1809 #define OBD_BRW_NOQUOTA 0x100
1810 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1811 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1812 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1813 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1814 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1816 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
1818 #define OST_MIN_PRECREATE 32
1819 #define OST_MAX_PRECREATE 20000
1822 struct ost_id ioo_oid
; /* object ID, if multi-obj BRW */
1823 __u32 ioo_max_brw
; /* low 16 bits were o_mode before 2.4,
1824 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1825 * high 16 bits in 2.4 and later */
1826 __u32 ioo_bufcnt
; /* number of niobufs for this object */
1829 #define IOOBJ_MAX_BRW_BITS 16
1830 #define IOOBJ_TYPE_MASK ((1U << IOOBJ_MAX_BRW_BITS) - 1)
1831 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1832 #define ioobj_max_brw_set(ioo, num) \
1833 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1835 extern void lustre_swab_obd_ioobj (struct obd_ioobj
*ioo
);
1837 /* multiple of 8 bytes => can array */
1838 struct niobuf_remote
{
1844 extern void lustre_swab_niobuf_remote (struct niobuf_remote
*nbr
);
1846 /* lock value block communicated between the filter and llite */
1848 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1849 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1850 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1851 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1852 #define OST_LVB_IS_ERR(blocks) \
1853 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1854 #define OST_LVB_SET_ERR(blocks, rc) \
1855 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1856 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1866 extern void lustre_swab_ost_lvb_v1(struct ost_lvb_v1
*lvb
);
1880 extern void lustre_swab_ost_lvb(struct ost_lvb
*lvb
);
1883 * lquota data structures
1886 #ifndef QUOTABLOCK_BITS
1887 #define QUOTABLOCK_BITS 10
1890 #ifndef QUOTABLOCK_SIZE
1891 #define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS)
1895 #define toqb(x) (((x) + QUOTABLOCK_SIZE - 1) >> QUOTABLOCK_BITS)
1898 /* The lquota_id structure is an union of all the possible identifier types that
1899 * can be used with quota, this includes:
1902 * - a FID which can be used for per-directory quota in the future */
1904 struct lu_fid qid_fid
; /* FID for per-directory quota */
1905 __u64 qid_uid
; /* user identifier */
1906 __u64 qid_gid
; /* group identifier */
1909 /* quotactl management */
1910 struct obd_quotactl
{
1912 __u32 qc_type
; /* see Q_* flag below */
1915 struct obd_dqinfo qc_dqinfo
;
1916 struct obd_dqblk qc_dqblk
;
1919 extern void lustre_swab_obd_quotactl(struct obd_quotactl
*q
);
1921 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1922 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1923 #define Q_GETOINFO 0x800102 /* get obd quota info */
1924 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1925 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1927 #define Q_COPY(out, in, member) (out)->member = (in)->member
1929 #define QCTL_COPY(out, in) \
1931 Q_COPY(out, in, qc_cmd); \
1932 Q_COPY(out, in, qc_type); \
1933 Q_COPY(out, in, qc_id); \
1934 Q_COPY(out, in, qc_stat); \
1935 Q_COPY(out, in, qc_dqinfo); \
1936 Q_COPY(out, in, qc_dqblk); \
1939 /* Body of quota request used for quota acquire/release RPCs between quota
1940 * master (aka QMT) and slaves (ak QSD). */
1942 struct lu_fid qb_fid
; /* FID of global index packing the pool ID
1943 * and type (data or metadata) as well as
1944 * the quota type (user or group). */
1945 union lquota_id qb_id
; /* uid or gid or directory FID */
1946 __u32 qb_flags
; /* see below */
1948 __u64 qb_count
; /* acquire/release count (kbytes/inodes) */
1949 __u64 qb_usage
; /* current slave usage (kbytes/inodes) */
1950 __u64 qb_slv_ver
; /* slave index file version */
1951 struct lustre_handle qb_lockh
; /* per-ID lock handle */
1952 struct lustre_handle qb_glb_lockh
; /* global lock handle */
1953 __u64 qb_padding1
[4];
1956 /* When the quota_body is used in the reply of quota global intent
1957 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1958 #define qb_slv_fid qb_fid
1959 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1961 #define qb_qunit qb_usage
1963 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1964 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1965 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1966 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1968 extern void lustre_swab_quota_body(struct quota_body
*b
);
1970 /* Quota types currently supported */
1972 LQUOTA_TYPE_USR
= 0x00, /* maps to USRQUOTA */
1973 LQUOTA_TYPE_GRP
= 0x01, /* maps to GRPQUOTA */
1977 /* There are 2 different resource types on which a quota limit can be enforced:
1978 * - inodes on the MDTs
1979 * - blocks on the OSTs */
1981 LQUOTA_RES_MD
= 0x01, /* skip 0 to avoid null oid in FID */
1982 LQUOTA_RES_DT
= 0x02,
1984 LQUOTA_FIRST_RES
= LQUOTA_RES_MD
1986 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1989 * Space accounting support
1990 * Format of an accounting record, providing disk usage information for a given
1993 struct lquota_acct_rec
{ /* 16 bytes */
1994 __u64 bspace
; /* current space in use */
1995 __u64 ispace
; /* current # inodes in use */
1999 * Global quota index support
2000 * Format of a global record, providing global quota settings for a given quota
2003 struct lquota_glb_rec
{ /* 32 bytes */
2004 __u64 qbr_hardlimit
; /* quota hard limit, in #inodes or kbytes */
2005 __u64 qbr_softlimit
; /* quota soft limit, in #inodes or kbytes */
2006 __u64 qbr_time
; /* grace time, in seconds */
2007 __u64 qbr_granted
; /* how much is granted to slaves, in #inodes or
2012 * Slave index support
2013 * Format of a slave record, recording how much space is granted to a given
2016 struct lquota_slv_rec
{ /* 8 bytes */
2017 __u64 qsr_granted
; /* space granted to the slave for the key=ID,
2018 * in #inodes or kbytes */
2021 /* Data structures associated with the quota locks */
2023 /* Glimpse descriptor used for the index & per-ID quota locks */
2024 struct ldlm_gl_lquota_desc
{
2025 union lquota_id gl_id
; /* quota ID subject to the glimpse */
2026 __u64 gl_flags
; /* see LQUOTA_FL* below */
2027 __u64 gl_ver
; /* new index version */
2028 __u64 gl_hardlimit
; /* new hardlimit or qunit value */
2029 __u64 gl_softlimit
; /* new softlimit */
2033 #define gl_qunit gl_hardlimit /* current qunit value used when
2034 * glimpsing per-ID quota locks */
2036 /* quota glimpse flags */
2037 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
2039 /* LVB used with quota (global and per-ID) locks */
2041 __u64 lvb_flags
; /* see LQUOTA_FL* above */
2042 __u64 lvb_id_may_rel
; /* space that might be released later */
2043 __u64 lvb_id_rel
; /* space released by the slave for this ID */
2044 __u64 lvb_id_qunit
; /* current qunit value */
2048 extern void lustre_swab_lquota_lvb(struct lquota_lvb
*lvb
);
2050 /* LVB used with global quota lock */
2051 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
2059 #define QUOTA_FIRST_OPC QUOTA_DQACQ
2068 MDS_GETATTR_NAME
= 34,
2073 MDS_DISCONNECT
= 39,
2079 MDS_DONE_WRITING
= 45,
2081 MDS_QUOTACHECK
= 47,
2084 MDS_SETXATTR
= 50, /* obsolete, now it's MDS_REINT op */
2088 MDS_HSM_STATE_GET
= 54,
2089 MDS_HSM_STATE_SET
= 55,
2090 MDS_HSM_ACTION
= 56,
2091 MDS_HSM_PROGRESS
= 57,
2092 MDS_HSM_REQUEST
= 58,
2093 MDS_HSM_CT_REGISTER
= 59,
2094 MDS_HSM_CT_UNREGISTER
= 60,
2095 MDS_SWAP_LAYOUTS
= 61,
2099 #define MDS_FIRST_OPC MDS_GETATTR
2102 /* opcodes for object update */
2108 #define UPDATE_FIRST_OPC UPDATE_OBJ
2125 } mds_reint_t
, mdt_reint_t
;
2127 extern void lustre_swab_generic_32s (__u32
*val
);
2129 /* the disposition of the intent outlines what was executed */
2130 #define DISP_IT_EXECD 0x00000001
2131 #define DISP_LOOKUP_EXECD 0x00000002
2132 #define DISP_LOOKUP_NEG 0x00000004
2133 #define DISP_LOOKUP_POS 0x00000008
2134 #define DISP_OPEN_CREATE 0x00000010
2135 #define DISP_OPEN_OPEN 0x00000020
2136 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
2137 #define DISP_ENQ_OPEN_REF 0x00800000
2138 #define DISP_ENQ_CREATE_REF 0x01000000
2139 #define DISP_OPEN_LOCK 0x02000000
2140 #define DISP_OPEN_LEASE 0x04000000
2141 #define DISP_OPEN_STRIPE 0x08000000
2143 /* INODE LOCK PARTS */
2144 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
2145 * was used to protect permission (mode,
2146 * owner, group etc) before 2.4. */
2147 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
2148 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
2149 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
2151 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
2152 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
2153 * Because for remote directories(in DNE), these locks will be granted by
2154 * different MDTs(different ldlm namespace).
2156 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
2157 * For Remote directory, the master MDT, where the remote directory is, will
2158 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
2159 * will grant LOOKUP_LOCK. */
2160 #define MDS_INODELOCK_PERM 0x000010
2161 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
2163 #define MDS_INODELOCK_MAXSHIFT 5
2164 /* This FULL lock is useful to take on unlink sort of operations */
2165 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
2167 extern void lustre_swab_ll_fid (struct ll_fid
*fid
);
2169 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
2170 * but was moved into name[1] along with the OID to avoid consuming the
2171 * name[2,3] fields that need to be used for the quota id (also a FID). */
2173 LUSTRE_RES_ID_SEQ_OFF
= 0,
2174 LUSTRE_RES_ID_VER_OID_OFF
= 1,
2175 LUSTRE_RES_ID_WAS_VER_OFF
= 2, /* see note above */
2176 LUSTRE_RES_ID_QUOTA_SEQ_OFF
= 2,
2177 LUSTRE_RES_ID_QUOTA_VER_OID_OFF
= 3,
2178 LUSTRE_RES_ID_HSH_OFF
= 3
2181 #define MDS_STATUS_CONN 1
2182 #define MDS_STATUS_LOV 2
2184 /* mdt_thread_info.mti_flags. */
2186 /* The flag indicates Size-on-MDS attributes are changed. */
2187 MF_SOM_CHANGE
= (1 << 0),
2188 /* Flags indicates an epoch opens or closes. */
2189 MF_EPOCH_OPEN
= (1 << 1),
2190 MF_EPOCH_CLOSE
= (1 << 2),
2191 MF_MDC_CANCEL_FID1
= (1 << 3),
2192 MF_MDC_CANCEL_FID2
= (1 << 4),
2193 MF_MDC_CANCEL_FID3
= (1 << 5),
2194 MF_MDC_CANCEL_FID4
= (1 << 6),
2195 /* There is a pending attribute update. */
2196 MF_SOM_AU
= (1 << 7),
2197 /* Cancel OST locks while getattr OST attributes. */
2198 MF_GETATTR_LOCK
= (1 << 8),
2199 MF_GET_MDT_IDX
= (1 << 9),
2202 #define MF_SOM_LOCAL_FLAGS (MF_SOM_CHANGE | MF_EPOCH_OPEN | MF_EPOCH_CLOSE)
2204 #define LUSTRE_BFLAG_UNCOMMITTED_WRITES 0x1
2206 /* these should be identical to their EXT4_*_FL counterparts, they are
2207 * redefined here only to avoid dragging in fs/ext4/ext4.h */
2208 #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */
2209 #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */
2210 #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */
2211 #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */
2212 #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
2214 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
2215 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
2216 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
2217 * the S_* flags are kernel-internal values that change between kernel
2218 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
2219 * See b=16526 for a full history. */
2220 static inline int ll_ext_to_inode_flags(int flags
)
2222 return (((flags
& LUSTRE_SYNC_FL
) ? S_SYNC
: 0) |
2223 ((flags
& LUSTRE_NOATIME_FL
) ? S_NOATIME
: 0) |
2224 ((flags
& LUSTRE_APPEND_FL
) ? S_APPEND
: 0) |
2225 #if defined(S_DIRSYNC)
2226 ((flags
& LUSTRE_DIRSYNC_FL
) ? S_DIRSYNC
: 0) |
2228 ((flags
& LUSTRE_IMMUTABLE_FL
) ? S_IMMUTABLE
: 0));
2231 static inline int ll_inode_to_ext_flags(int iflags
)
2233 return (((iflags
& S_SYNC
) ? LUSTRE_SYNC_FL
: 0) |
2234 ((iflags
& S_NOATIME
) ? LUSTRE_NOATIME_FL
: 0) |
2235 ((iflags
& S_APPEND
) ? LUSTRE_APPEND_FL
: 0) |
2236 #if defined(S_DIRSYNC)
2237 ((iflags
& S_DIRSYNC
) ? LUSTRE_DIRSYNC_FL
: 0) |
2239 ((iflags
& S_IMMUTABLE
) ? LUSTRE_IMMUTABLE_FL
: 0));
2242 /* 64 possible states */
2243 enum md_transient_state
{
2244 MS_RESTORE
= (1 << 0), /* restore is running */
2250 struct lustre_handle handle
;
2252 __u64 size
; /* Offset, in the case of MDS_READPAGE */
2256 __u64 blocks
; /* XID, in the case of MDS_READPAGE */
2258 __u64 t_state
; /* transient file state defined in
2259 * enum md_transient_state
2260 * was "ino" until 2.4.0 */
2267 __u32 flags
; /* from vfs for pin/unpin, LUSTRE_BFLAG close */
2269 __u32 nlink
; /* #bytes to read in the case of MDS_READPAGE */
2270 __u32 unused2
; /* was "generation" until 2.4.0 */
2275 __u32 max_cookiesize
;
2276 __u32 uid_h
; /* high 32-bits of uid, for FUID */
2277 __u32 gid_h
; /* high 32-bits of gid, for FUID */
2278 __u32 padding_5
; /* also fix lustre_swab_mdt_body */
2286 extern void lustre_swab_mdt_body (struct mdt_body
*b
);
2288 struct mdt_ioepoch
{
2289 struct lustre_handle handle
;
2295 extern void lustre_swab_mdt_ioepoch (struct mdt_ioepoch
*b
);
2297 /* permissions for md_perm.mp_perm */
2299 CFS_SETUID_PERM
= 0x01,
2300 CFS_SETGID_PERM
= 0x02,
2301 CFS_SETGRP_PERM
= 0x04,
2302 CFS_RMTACL_PERM
= 0x08,
2303 CFS_RMTOWN_PERM
= 0x10
2306 /* inode access permission for remote user, the inode info are omitted,
2307 * for client knows them. */
2308 struct mdt_remote_perm
{
2315 __u32 rp_access_perm
; /* MAY_READ/WRITE/EXEC */
2319 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm
*p
);
2321 struct mdt_rec_setattr
{
2331 __u32 sa_padding_1_h
;
2332 struct lu_fid sa_fid
;
2341 __u32 sa_attr_flags
;
2343 __u32 sa_bias
; /* some operation flags */
2349 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr
*sa
);
2352 * Attribute flags used in mdt_rec_setattr::sa_valid.
2353 * The kernel's #defines for ATTR_* should not be used over the network
2354 * since the client and MDS may run different kernels (see bug 13828)
2355 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
2357 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
2358 #define MDS_ATTR_UID 0x2ULL /* = 2 */
2359 #define MDS_ATTR_GID 0x4ULL /* = 4 */
2360 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
2361 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
2362 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
2363 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
2364 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
2365 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
2366 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
2367 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
2368 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
2369 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
2370 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
2371 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
2372 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
2375 #define FMODE_READ 00000001
2376 #define FMODE_WRITE 00000002
2379 #define MDS_FMODE_CLOSED 00000000
2380 #define MDS_FMODE_EXEC 00000004
2381 /* IO Epoch is opened on a closed file. */
2382 #define MDS_FMODE_EPOCH 01000000
2383 /* IO Epoch is opened on a file truncate. */
2384 #define MDS_FMODE_TRUNC 02000000
2385 /* Size-on-MDS Attribute Update is pending. */
2386 #define MDS_FMODE_SOM 04000000
2388 #define MDS_OPEN_CREATED 00000010
2389 #define MDS_OPEN_CROSS 00000020
2391 #define MDS_OPEN_CREAT 00000100
2392 #define MDS_OPEN_EXCL 00000200
2393 #define MDS_OPEN_TRUNC 00001000
2394 #define MDS_OPEN_APPEND 00002000
2395 #define MDS_OPEN_SYNC 00010000
2396 #define MDS_OPEN_DIRECTORY 00200000
2398 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
2399 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
2400 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
2401 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
2402 * We do not support JOIN FILE
2403 * anymore, reserve this flags
2404 * just for preventing such bit
2407 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
2408 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
2409 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
2410 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
2411 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
2413 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
2415 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
2416 * delegation, succeed if it's not
2417 * being opened with conflict mode.
2419 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
2421 /* permission for create non-directory file */
2422 #define MAY_CREATE (1 << 7)
2423 /* permission for create directory file */
2424 #define MAY_LINK (1 << 8)
2425 /* permission for delete from the directory */
2426 #define MAY_UNLINK (1 << 9)
2427 /* source's permission for rename */
2428 #define MAY_RENAME_SRC (1 << 10)
2429 /* target's permission for rename */
2430 #define MAY_RENAME_TAR (1 << 11)
2431 /* part (parent's) VTX permission check */
2432 #define MAY_VTX_PART (1 << 12)
2433 /* full VTX permission check */
2434 #define MAY_VTX_FULL (1 << 13)
2435 /* lfs rgetfacl permission check */
2436 #define MAY_RGETFACL (1 << 14)
2439 MDS_CHECK_SPLIT
= 1 << 0,
2440 MDS_CROSS_REF
= 1 << 1,
2441 MDS_VTX_BYPASS
= 1 << 2,
2442 MDS_PERM_BYPASS
= 1 << 3,
2444 MDS_QUOTA_IGNORE
= 1 << 5,
2445 MDS_CLOSE_CLEANUP
= 1 << 6,
2446 MDS_KEEP_ORPHAN
= 1 << 7,
2447 MDS_RECOV_OPEN
= 1 << 8,
2448 MDS_DATA_MODIFIED
= 1 << 9,
2449 MDS_CREATE_VOLATILE
= 1 << 10,
2450 MDS_OWNEROVERRIDE
= 1 << 11,
2451 MDS_HSM_RELEASE
= 1 << 12,
2454 /* instance of mdt_reint_rec */
2455 struct mdt_rec_create
{
2463 __u32 cr_suppgid1_h
;
2465 __u32 cr_suppgid2_h
;
2466 struct lu_fid cr_fid1
;
2467 struct lu_fid cr_fid2
;
2468 struct lustre_handle cr_old_handle
; /* handle in case of open replay */
2472 __u64 cr_padding_1
; /* rr_blocks */
2475 /* use of helpers set/get_mrc_cr_flags() is needed to access
2476 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2477 * extend cr_flags size without breaking 1.8 compat */
2478 __u32 cr_flags_l
; /* for use with open, low 32 bits */
2479 __u32 cr_flags_h
; /* for use with open, high 32 bits */
2480 __u32 cr_umask
; /* umask for create */
2481 __u32 cr_padding_4
; /* rr_padding_4 */
2484 static inline void set_mrc_cr_flags(struct mdt_rec_create
*mrc
, __u64 flags
)
2486 mrc
->cr_flags_l
= (__u32
)(flags
& 0xFFFFFFFFUll
);
2487 mrc
->cr_flags_h
= (__u32
)(flags
>> 32);
2490 static inline __u64
get_mrc_cr_flags(struct mdt_rec_create
*mrc
)
2492 return ((__u64
)(mrc
->cr_flags_l
) | ((__u64
)mrc
->cr_flags_h
<< 32));
2495 /* instance of mdt_reint_rec */
2496 struct mdt_rec_link
{
2504 __u32 lk_suppgid1_h
;
2506 __u32 lk_suppgid2_h
;
2507 struct lu_fid lk_fid1
;
2508 struct lu_fid lk_fid2
;
2510 __u64 lk_padding_1
; /* rr_atime */
2511 __u64 lk_padding_2
; /* rr_ctime */
2512 __u64 lk_padding_3
; /* rr_size */
2513 __u64 lk_padding_4
; /* rr_blocks */
2515 __u32 lk_padding_5
; /* rr_mode */
2516 __u32 lk_padding_6
; /* rr_flags */
2517 __u32 lk_padding_7
; /* rr_padding_2 */
2518 __u32 lk_padding_8
; /* rr_padding_3 */
2519 __u32 lk_padding_9
; /* rr_padding_4 */
2522 /* instance of mdt_reint_rec */
2523 struct mdt_rec_unlink
{
2531 __u32 ul_suppgid1_h
;
2533 __u32 ul_suppgid2_h
;
2534 struct lu_fid ul_fid1
;
2535 struct lu_fid ul_fid2
;
2537 __u64 ul_padding_2
; /* rr_atime */
2538 __u64 ul_padding_3
; /* rr_ctime */
2539 __u64 ul_padding_4
; /* rr_size */
2540 __u64 ul_padding_5
; /* rr_blocks */
2543 __u32 ul_padding_6
; /* rr_flags */
2544 __u32 ul_padding_7
; /* rr_padding_2 */
2545 __u32 ul_padding_8
; /* rr_padding_3 */
2546 __u32 ul_padding_9
; /* rr_padding_4 */
2549 /* instance of mdt_reint_rec */
2550 struct mdt_rec_rename
{
2558 __u32 rn_suppgid1_h
;
2560 __u32 rn_suppgid2_h
;
2561 struct lu_fid rn_fid1
;
2562 struct lu_fid rn_fid2
;
2564 __u64 rn_padding_1
; /* rr_atime */
2565 __u64 rn_padding_2
; /* rr_ctime */
2566 __u64 rn_padding_3
; /* rr_size */
2567 __u64 rn_padding_4
; /* rr_blocks */
2568 __u32 rn_bias
; /* some operation flags */
2569 __u32 rn_mode
; /* cross-ref rename has mode */
2570 __u32 rn_padding_5
; /* rr_flags */
2571 __u32 rn_padding_6
; /* rr_padding_2 */
2572 __u32 rn_padding_7
; /* rr_padding_3 */
2573 __u32 rn_padding_8
; /* rr_padding_4 */
2576 /* instance of mdt_reint_rec */
2577 struct mdt_rec_setxattr
{
2585 __u32 sx_suppgid1_h
;
2587 __u32 sx_suppgid2_h
;
2588 struct lu_fid sx_fid
;
2589 __u64 sx_padding_1
; /* These three are rr_fid2 */
2594 __u64 sx_padding_5
; /* rr_ctime */
2595 __u64 sx_padding_6
; /* rr_size */
2596 __u64 sx_padding_7
; /* rr_blocks */
2599 __u32 sx_padding_8
; /* rr_flags */
2600 __u32 sx_padding_9
; /* rr_padding_2 */
2601 __u32 sx_padding_10
; /* rr_padding_3 */
2602 __u32 sx_padding_11
; /* rr_padding_4 */
2606 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2607 * Do NOT change the size of various members, otherwise the value
2608 * will be broken in lustre_swab_mdt_rec_reint().
2610 * If you add new members in other mdt_reint_xxx structures and need to use the
2611 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2613 struct mdt_rec_reint
{
2621 __u32 rr_suppgid1_h
;
2623 __u32 rr_suppgid2_h
;
2624 struct lu_fid rr_fid1
;
2625 struct lu_fid rr_fid2
;
2636 __u32 rr_padding_4
; /* also fix lustre_swab_mdt_rec_reint */
2639 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint
*rr
);
2642 __u32 ld_tgt_count
; /* how many MDS's */
2643 __u32 ld_active_tgt_count
; /* how many active */
2644 __u32 ld_default_stripe_count
; /* how many objects are used */
2645 __u32 ld_pattern
; /* default MEA_MAGIC_* */
2646 __u64 ld_default_hash_size
;
2647 __u64 ld_padding_1
; /* also fix lustre_swab_lmv_desc */
2648 __u32 ld_padding_2
; /* also fix lustre_swab_lmv_desc */
2649 __u32 ld_qos_maxage
; /* in second */
2650 __u32 ld_padding_3
; /* also fix lustre_swab_lmv_desc */
2651 __u32 ld_padding_4
; /* also fix lustre_swab_lmv_desc */
2652 struct obd_uuid ld_uuid
;
2655 extern void lustre_swab_lmv_desc (struct lmv_desc
*ld
);
2657 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
2658 struct lmv_stripe_md
{
2663 char mea_pool_name
[LOV_MAXPOOLNAME
];
2664 struct lu_fid mea_ids
[0];
2667 extern void lustre_swab_lmv_stripe_md(struct lmv_stripe_md
*mea
);
2669 /* lmv structures */
2670 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
2671 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
2672 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
2674 #define MAX_HASH_SIZE_32 0x7fffffffUL
2675 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
2676 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
2681 FLD_FIRST_OPC
= FLD_QUERY
2687 SEQ_FIRST_OPC
= SEQ_QUERY
2691 SEQ_ALLOC_SUPER
= 0,
2696 * LOV data structures
2699 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2700 /* The size of the buffer the lov/mdc reserves for the
2701 * array of UUIDs returned by the MDS. With the current
2702 * protocol, this will limit the max number of OSTs per LOV */
2704 #define LOV_DESC_MAGIC 0xB0CCDE5C
2705 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2706 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2708 /* LOV settings descriptor (should only contain static info) */
2710 __u32 ld_tgt_count
; /* how many OBD's */
2711 __u32 ld_active_tgt_count
; /* how many active */
2712 __u32 ld_default_stripe_count
; /* how many objects are used */
2713 __u32 ld_pattern
; /* default PATTERN_RAID0 */
2714 __u64 ld_default_stripe_size
; /* in bytes */
2715 __u64 ld_default_stripe_offset
; /* in bytes */
2716 __u32 ld_padding_0
; /* unused */
2717 __u32 ld_qos_maxage
; /* in second */
2718 __u32 ld_padding_1
; /* also fix lustre_swab_lov_desc */
2719 __u32 ld_padding_2
; /* also fix lustre_swab_lov_desc */
2720 struct obd_uuid ld_uuid
;
2723 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2725 extern void lustre_swab_lov_desc (struct lov_desc
*ld
);
2730 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2735 LDLM_BL_CALLBACK
= 104,
2736 LDLM_CP_CALLBACK
= 105,
2737 LDLM_GL_CALLBACK
= 106,
2738 LDLM_SET_INFO
= 107,
2741 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2743 #define RES_NAME_SIZE 4
2744 struct ldlm_res_id
{
2745 __u64 name
[RES_NAME_SIZE
];
2748 #define DLDLMRES "[%#llx:%#llx:%#llx].%llx"
2749 #define PLDLMRES(res) (res)->lr_name.name[0], (res)->lr_name.name[1], \
2750 (res)->lr_name.name[2], (res)->lr_name.name[3]
2752 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id
*id
);
2754 static inline int ldlm_res_eq(const struct ldlm_res_id
*res0
,
2755 const struct ldlm_res_id
*res1
)
2757 return !memcmp(res0
, res1
, sizeof(*res0
));
2774 #define LCK_MODE_NUM 8
2784 #define LDLM_MIN_TYPE LDLM_PLAIN
2786 struct ldlm_extent
{
2792 static inline int ldlm_extent_overlap(struct ldlm_extent
*ex1
,
2793 struct ldlm_extent
*ex2
)
2795 return (ex1
->start
<= ex2
->end
) && (ex2
->start
<= ex1
->end
);
2798 /* check if @ex1 contains @ex2 */
2799 static inline int ldlm_extent_contain(struct ldlm_extent
*ex1
,
2800 struct ldlm_extent
*ex2
)
2802 return (ex1
->start
<= ex2
->start
) && (ex1
->end
>= ex2
->end
);
2805 struct ldlm_inodebits
{
2809 struct ldlm_flock_wire
{
2817 /* it's important that the fields of the ldlm_extent structure match
2818 * the first fields of the ldlm_flock structure because there is only
2819 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2820 * this ever changes we will need to swab the union differently based
2821 * on the resource type. */
2824 struct ldlm_extent l_extent
;
2825 struct ldlm_flock_wire l_flock
;
2826 struct ldlm_inodebits l_inodebits
;
2827 } ldlm_wire_policy_data_t
;
2829 extern void lustre_swab_ldlm_policy_data (ldlm_wire_policy_data_t
*d
);
2831 union ldlm_gl_desc
{
2832 struct ldlm_gl_lquota_desc lquota_desc
;
2835 extern void lustre_swab_gl_desc(union ldlm_gl_desc
*);
2837 struct ldlm_intent
{
2841 extern void lustre_swab_ldlm_intent (struct ldlm_intent
*i
);
2843 struct ldlm_resource_desc
{
2844 ldlm_type_t lr_type
;
2845 __u32 lr_padding
; /* also fix lustre_swab_ldlm_resource_desc */
2846 struct ldlm_res_id lr_name
;
2849 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc
*r
);
2851 struct ldlm_lock_desc
{
2852 struct ldlm_resource_desc l_resource
;
2853 ldlm_mode_t l_req_mode
;
2854 ldlm_mode_t l_granted_mode
;
2855 ldlm_wire_policy_data_t l_policy_data
;
2858 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc
*l
);
2860 #define LDLM_LOCKREQ_HANDLES 2
2861 #define LDLM_ENQUEUE_CANCEL_OFF 1
2863 struct ldlm_request
{
2866 struct ldlm_lock_desc lock_desc
;
2867 struct lustre_handle lock_handle
[LDLM_LOCKREQ_HANDLES
];
2870 extern void lustre_swab_ldlm_request (struct ldlm_request
*rq
);
2872 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
2873 * Otherwise, 2 are available. */
2874 #define ldlm_request_bufsize(count,type) \
2876 int _avail = LDLM_LOCKREQ_HANDLES; \
2877 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
2878 sizeof(struct ldlm_request) + \
2879 (count > _avail ? count - _avail : 0) * \
2880 sizeof(struct lustre_handle); \
2885 __u32 lock_padding
; /* also fix lustre_swab_ldlm_reply */
2886 struct ldlm_lock_desc lock_desc
;
2887 struct lustre_handle lock_handle
;
2888 __u64 lock_policy_res1
;
2889 __u64 lock_policy_res2
;
2892 extern void lustre_swab_ldlm_reply (struct ldlm_reply
*r
);
2894 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2895 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2898 * Opcodes for mountconf (mgs and mgc)
2903 MGS_EXCEPTION
, /* node died, etc. */
2904 MGS_TARGET_REG
, /* whenever target starts up */
2910 #define MGS_FIRST_OPC MGS_CONNECT
2912 #define MGS_PARAM_MAXLEN 1024
2913 #define KEY_SET_INFO "set_info"
2915 struct mgs_send_param
{
2916 char mgs_param
[MGS_PARAM_MAXLEN
];
2919 /* We pass this info to the MGS so it can write config logs */
2920 #define MTI_NAME_MAXLEN 64
2921 #define MTI_PARAM_MAXLEN 4096
2922 #define MTI_NIDS_MAX 32
2923 struct mgs_target_info
{
2924 __u32 mti_lustre_ver
;
2925 __u32 mti_stripe_index
;
2926 __u32 mti_config_ver
;
2928 __u32 mti_nid_count
;
2929 __u32 mti_instance
; /* Running instance of target */
2930 char mti_fsname
[MTI_NAME_MAXLEN
];
2931 char mti_svname
[MTI_NAME_MAXLEN
];
2932 char mti_uuid
[sizeof(struct obd_uuid
)];
2933 __u64 mti_nids
[MTI_NIDS_MAX
]; /* host nids (lnet_nid_t)*/
2934 char mti_params
[MTI_PARAM_MAXLEN
];
2936 extern void lustre_swab_mgs_target_info(struct mgs_target_info
*oinfo
);
2938 struct mgs_nidtbl_entry
{
2939 __u64 mne_version
; /* table version of this entry */
2940 __u32 mne_instance
; /* target instance # */
2941 __u32 mne_index
; /* target index */
2942 __u32 mne_length
; /* length of this entry - by bytes */
2943 __u8 mne_type
; /* target type LDD_F_SV_TYPE_OST/MDT */
2944 __u8 mne_nid_type
; /* type of nid(mbz). for ipv6. */
2945 __u8 mne_nid_size
; /* size of each NID, by bytes */
2946 __u8 mne_nid_count
; /* # of NIDs in buffer */
2948 lnet_nid_t nids
[0]; /* variable size buffer for NIDs. */
2951 extern void lustre_swab_mgs_nidtbl_entry(struct mgs_nidtbl_entry
*oinfo
);
2953 struct mgs_config_body
{
2954 char mcb_name
[MTI_NAME_MAXLEN
]; /* logname */
2955 __u64 mcb_offset
; /* next index of config log to request */
2956 __u16 mcb_type
; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2958 __u8 mcb_bits
; /* bits unit size of config log */
2959 __u32 mcb_units
; /* # of units for bulk transfer */
2961 extern void lustre_swab_mgs_config_body(struct mgs_config_body
*body
);
2963 struct mgs_config_res
{
2964 __u64 mcr_offset
; /* index of last config log */
2965 __u64 mcr_size
; /* size of the log */
2967 extern void lustre_swab_mgs_config_res(struct mgs_config_res
*body
);
2969 /* Config marker flags (in config log) */
2970 #define CM_START 0x01
2972 #define CM_SKIP 0x04
2973 #define CM_UPGRADE146 0x08
2974 #define CM_EXCLUDE 0x10
2975 #define CM_START_SKIP (CM_START | CM_SKIP)
2978 __u32 cm_step
; /* aka config version */
2980 __u32 cm_vers
; /* lustre release version number */
2981 __u32 cm_padding
; /* 64 bit align */
2982 obd_time cm_createtime
; /*when this record was first created */
2983 obd_time cm_canceltime
; /*when this record is no longer valid*/
2984 char cm_tgtname
[MTI_NAME_MAXLEN
];
2985 char cm_comment
[MTI_NAME_MAXLEN
];
2988 extern void lustre_swab_cfg_marker(struct cfg_marker
*marker
,
2989 int swab
, int size
);
2992 * Opcodes for multiple servers.
3002 #define OBD_FIRST_OPC OBD_PING
3004 /* catalog of log objects */
3006 /** Identifier for a single log object */
3008 struct ost_id lgl_oi
;
3010 } __attribute__((packed
));
3012 /** Records written to the CATALOGS list */
3013 #define CATLIST "CATALOGS"
3015 struct llog_logid lci_logid
;
3019 } __attribute__((packed
));
3021 /* Log data record types - there is no specific reason that these need to
3022 * be related to the RPC opcodes, but no reason not to (may be handy later?)
3024 #define LLOG_OP_MAGIC 0x10600000
3025 #define LLOG_OP_MASK 0xfff00000
3028 LLOG_PAD_MAGIC
= LLOG_OP_MAGIC
| 0x00000,
3029 OST_SZ_REC
= LLOG_OP_MAGIC
| 0x00f00,
3030 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
3031 MDS_UNLINK_REC
= LLOG_OP_MAGIC
| 0x10000 | (MDS_REINT
<< 8) |
3032 REINT_UNLINK
, /* obsolete after 2.5.0 */
3033 MDS_UNLINK64_REC
= LLOG_OP_MAGIC
| 0x90000 | (MDS_REINT
<< 8) |
3035 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
3036 MDS_SETATTR64_REC
= LLOG_OP_MAGIC
| 0x90000 | (MDS_REINT
<< 8) |
3038 OBD_CFG_REC
= LLOG_OP_MAGIC
| 0x20000,
3039 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
3040 LLOG_GEN_REC
= LLOG_OP_MAGIC
| 0x40000,
3041 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
3042 CHANGELOG_REC
= LLOG_OP_MAGIC
| 0x60000,
3043 CHANGELOG_USER_REC
= LLOG_OP_MAGIC
| 0x70000,
3044 HSM_AGENT_REC
= LLOG_OP_MAGIC
| 0x80000,
3045 LLOG_HDR_MAGIC
= LLOG_OP_MAGIC
| 0x45539,
3046 LLOG_LOGID_MAGIC
= LLOG_OP_MAGIC
| 0x4553b,
3049 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
3050 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
3052 /** Log record header - stored in little endian order.
3053 * Each record must start with this struct, end with a llog_rec_tail,
3054 * and be a multiple of 256 bits in size.
3056 struct llog_rec_hdr
{
3063 struct llog_rec_tail
{
3068 /* Where data follow just after header */
3069 #define REC_DATA(ptr) \
3070 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
3072 #define REC_DATA_LEN(rec) \
3073 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
3074 sizeof(struct llog_rec_tail))
3076 struct llog_logid_rec
{
3077 struct llog_rec_hdr lid_hdr
;
3078 struct llog_logid lid_id
;
3082 struct llog_rec_tail lid_tail
;
3083 } __attribute__((packed
));
3085 struct llog_unlink_rec
{
3086 struct llog_rec_hdr lur_hdr
;
3089 obd_count lur_count
;
3090 struct llog_rec_tail lur_tail
;
3091 } __attribute__((packed
));
3093 struct llog_unlink64_rec
{
3094 struct llog_rec_hdr lur_hdr
;
3095 struct lu_fid lur_fid
;
3096 obd_count lur_count
; /* to destroy the lost precreated */
3100 struct llog_rec_tail lur_tail
;
3101 } __attribute__((packed
));
3103 struct llog_setattr64_rec
{
3104 struct llog_rec_hdr lsr_hdr
;
3105 struct ost_id lsr_oi
;
3111 struct llog_rec_tail lsr_tail
;
3112 } __attribute__((packed
));
3114 struct llog_size_change_rec
{
3115 struct llog_rec_hdr lsc_hdr
;
3116 struct ll_fid lsc_fid
;
3121 struct llog_rec_tail lsc_tail
;
3122 } __attribute__((packed
));
3124 #define CHANGELOG_MAGIC 0xca103000
3126 /** \a changelog_rec_type's that can't be masked */
3127 #define CHANGELOG_MINMASK (1 << CL_MARK)
3128 /** bits covering all \a changelog_rec_type's */
3129 #define CHANGELOG_ALLMASK 0XFFFFFFFF
3130 /** default \a changelog_rec_type mask */
3131 #define CHANGELOG_DEFMASK CHANGELOG_ALLMASK & ~(1 << CL_ATIME | 1 << CL_CLOSE)
3133 /* changelog llog name, needed by client replicators */
3134 #define CHANGELOG_CATALOG "changelog_catalog"
3136 struct changelog_setinfo
{
3139 } __attribute__((packed
));
3141 /** changelog record */
3142 struct llog_changelog_rec
{
3143 struct llog_rec_hdr cr_hdr
;
3144 struct changelog_rec cr
;
3145 struct llog_rec_tail cr_tail
; /**< for_sizezof_only */
3146 } __attribute__((packed
));
3148 struct llog_changelog_ext_rec
{
3149 struct llog_rec_hdr cr_hdr
;
3150 struct changelog_ext_rec cr
;
3151 struct llog_rec_tail cr_tail
; /**< for_sizezof_only */
3152 } __attribute__((packed
));
3154 #define CHANGELOG_USER_PREFIX "cl"
3156 struct llog_changelog_user_rec
{
3157 struct llog_rec_hdr cur_hdr
;
3161 struct llog_rec_tail cur_tail
;
3162 } __attribute__((packed
));
3164 enum agent_req_status
{
3172 static inline char *agent_req_status2name(enum agent_req_status ars
)
3190 static inline bool agent_req_in_final_state(enum agent_req_status ars
)
3192 return ((ars
== ARS_SUCCEED
) || (ars
== ARS_FAILED
) ||
3193 (ars
== ARS_CANCELED
));
3196 struct llog_agent_req_rec
{
3197 struct llog_rec_hdr arr_hdr
; /**< record header */
3198 __u32 arr_status
; /**< status of the request */
3200 * agent_req_status */
3201 __u32 arr_archive_id
; /**< backend archive number */
3202 __u64 arr_flags
; /**< req flags */
3203 __u64 arr_compound_id
; /**< compound cookie */
3204 __u64 arr_req_create
; /**< req. creation time */
3205 __u64 arr_req_change
; /**< req. status change time */
3206 struct hsm_action_item arr_hai
; /**< req. to the agent */
3207 struct llog_rec_tail arr_tail
; /**< record tail for_sizezof_only */
3208 } __attribute__((packed
));
3210 /* Old llog gen for compatibility */
3214 } __attribute__((packed
));
3216 struct llog_gen_rec
{
3217 struct llog_rec_hdr lgr_hdr
;
3218 struct llog_gen lgr_gen
;
3222 struct llog_rec_tail lgr_tail
;
3225 /* On-disk header structure of each log object, stored in little endian order */
3226 #define LLOG_CHUNK_SIZE 8192
3227 #define LLOG_HEADER_SIZE (96)
3228 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
3230 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
3232 /* flags for the logs */
3234 LLOG_F_ZAP_WHEN_EMPTY
= 0x1,
3235 LLOG_F_IS_CAT
= 0x2,
3236 LLOG_F_IS_PLAIN
= 0x4,
3239 struct llog_log_hdr
{
3240 struct llog_rec_hdr llh_hdr
;
3241 obd_time llh_timestamp
;
3243 __u32 llh_bitmap_offset
;
3247 /* for a catalog the first plain slot is next to it */
3248 struct obd_uuid llh_tgtuuid
;
3249 __u32 llh_reserved
[LLOG_HEADER_SIZE
/sizeof(__u32
) - 23];
3250 __u32 llh_bitmap
[LLOG_BITMAP_BYTES
/sizeof(__u32
)];
3251 struct llog_rec_tail llh_tail
;
3252 } __attribute__((packed
));
3254 #define LLOG_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
3255 llh->llh_bitmap_offset - \
3256 sizeof(llh->llh_tail)) * 8)
3258 /** log cookies are used to reference a specific log file and a record therein */
3259 struct llog_cookie
{
3260 struct llog_logid lgc_lgl
;
3264 } __attribute__((packed
));
3266 /** llog protocol */
3267 enum llogd_rpc_ops
{
3268 LLOG_ORIGIN_HANDLE_CREATE
= 501,
3269 LLOG_ORIGIN_HANDLE_NEXT_BLOCK
= 502,
3270 LLOG_ORIGIN_HANDLE_READ_HEADER
= 503,
3271 LLOG_ORIGIN_HANDLE_WRITE_REC
= 504,
3272 LLOG_ORIGIN_HANDLE_CLOSE
= 505,
3273 LLOG_ORIGIN_CONNECT
= 506,
3274 LLOG_CATINFO
= 507, /* deprecated */
3275 LLOG_ORIGIN_HANDLE_PREV_BLOCK
= 508,
3276 LLOG_ORIGIN_HANDLE_DESTROY
= 509, /* for destroy llog object*/
3278 LLOG_FIRST_OPC
= LLOG_ORIGIN_HANDLE_CREATE
3282 struct llog_logid lgd_logid
;
3284 __u32 lgd_llh_flags
;
3286 __u32 lgd_saved_index
;
3288 __u64 lgd_cur_offset
;
3289 } __attribute__((packed
));
3291 struct llogd_conn_body
{
3292 struct llog_gen lgdc_gen
;
3293 struct llog_logid lgdc_logid
;
3294 __u32 lgdc_ctxt_idx
;
3295 } __attribute__((packed
));
3297 /* Note: 64-bit types are 64-bit aligned in structure */
3299 obd_valid o_valid
; /* hot fields in this obdo */
3301 obd_id o_parent_seq
;
3302 obd_size o_size
; /* o_size-o_blocks == ost_lvb */
3306 obd_blocks o_blocks
; /* brw: cli sent cached bytes */
3309 /* 32-bit fields start here: keep an even number of them via padding */
3310 obd_blksize o_blksize
; /* optimal IO blocksize */
3311 obd_mode o_mode
; /* brw: cli sent cache remain */
3315 obd_count o_nlink
; /* brw: checksum */
3316 obd_count o_parent_oid
;
3317 obd_count o_misc
; /* brw: o_dropped */
3319 __u64 o_ioepoch
; /* epoch in ost writes */
3320 __u32 o_stripe_idx
; /* holds stripe idx */
3322 struct lustre_handle o_handle
; /* brw: lock handle to prolong
3324 struct llog_cookie o_lcookie
; /* destroy: unlink cookie from
3329 __u64 o_data_version
; /* getattr: sum of iversion for
3331 * brw: grant space consumed on
3332 * the client for the write */
3338 #define o_dirty o_blocks
3339 #define o_undirty o_mode
3340 #define o_dropped o_misc
3341 #define o_cksum o_nlink
3342 #define o_grant_used o_data_version
3344 static inline void lustre_set_wire_obdo(struct obd_connect_data
*ocd
,
3346 const struct obdo
*lobdo
)
3349 wobdo
->o_flags
&= ~OBD_FL_LOCAL_MASK
;
3353 if (unlikely(!(ocd
->ocd_connect_flags
& OBD_CONNECT_FID
)) &&
3354 fid_seq_is_echo(ostid_seq(&lobdo
->o_oi
))) {
3355 /* Currently OBD_FL_OSTID will only be used when 2.4 echo
3356 * client communicate with pre-2.4 server */
3357 wobdo
->o_oi
.oi
.oi_id
= fid_oid(&lobdo
->o_oi
.oi_fid
);
3358 wobdo
->o_oi
.oi
.oi_seq
= fid_seq(&lobdo
->o_oi
.oi_fid
);
3362 static inline void lustre_get_wire_obdo(struct obd_connect_data
*ocd
,
3364 const struct obdo
*wobdo
)
3366 obd_flag local_flags
= 0;
3368 if (lobdo
->o_valid
& OBD_MD_FLFLAGS
)
3369 local_flags
= lobdo
->o_flags
& OBD_FL_LOCAL_MASK
;
3372 if (local_flags
!= 0) {
3373 lobdo
->o_valid
|= OBD_MD_FLFLAGS
;
3374 lobdo
->o_flags
&= ~OBD_FL_LOCAL_MASK
;
3375 lobdo
->o_flags
|= local_flags
;
3380 if (unlikely(!(ocd
->ocd_connect_flags
& OBD_CONNECT_FID
)) &&
3381 fid_seq_is_echo(wobdo
->o_oi
.oi
.oi_seq
)) {
3383 lobdo
->o_oi
.oi_fid
.f_seq
= wobdo
->o_oi
.oi
.oi_seq
;
3384 lobdo
->o_oi
.oi_fid
.f_oid
= wobdo
->o_oi
.oi
.oi_id
;
3385 lobdo
->o_oi
.oi_fid
.f_ver
= 0;
3389 extern void lustre_swab_obdo (struct obdo
*o
);
3391 /* request structure for OST's */
3396 /* Key for FIEMAP to be used in get_info calls */
3397 struct ll_fiemap_info_key
{
3400 struct ll_user_fiemap fiemap
;
3403 extern void lustre_swab_ost_body (struct ost_body
*b
);
3404 extern void lustre_swab_ost_last_id(obd_id
*id
);
3405 extern void lustre_swab_fiemap(struct ll_user_fiemap
*fiemap
);
3407 extern void lustre_swab_lov_user_md_v1(struct lov_user_md_v1
*lum
);
3408 extern void lustre_swab_lov_user_md_v3(struct lov_user_md_v3
*lum
);
3409 extern void lustre_swab_lov_user_md_objects(struct lov_user_ost_data
*lod
,
3411 extern void lustre_swab_lov_mds_md(struct lov_mds_md
*lmm
);
3414 extern void lustre_swab_llogd_body (struct llogd_body
*d
);
3415 extern void lustre_swab_llog_hdr (struct llog_log_hdr
*h
);
3416 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body
*d
);
3417 extern void lustre_swab_llog_rec(struct llog_rec_hdr
*rec
);
3418 extern void lustre_swab_llog_id(struct llog_logid
*lid
);
3421 extern void lustre_swab_lustre_cfg(struct lustre_cfg
*lcfg
);
3423 /* Functions for dumping PTLRPC fields */
3424 void dump_rniobuf(struct niobuf_remote
*rnb
);
3425 void dump_ioo(struct obd_ioobj
*nb
);
3426 void dump_obdo(struct obdo
*oa
);
3427 void dump_ost_body(struct ost_body
*ob
);
3428 void dump_rcs(__u32
*rc
);
3430 #define IDX_INFO_MAGIC 0x3D37CC37
3432 /* Index file transfer through the network. The server serializes the index into
3433 * a byte stream which is sent to the client via a bulk transfer */
3437 /* reply: see idx_info_flags below */
3440 /* request & reply: number of lu_idxpage (to be) transferred */
3444 /* request: requested attributes passed down to the iterator API */
3447 /* request & reply: index file identifier (FID) */
3448 struct lu_fid ii_fid
;
3450 /* reply: version of the index file before starting to walk the index.
3451 * Please note that the version can be modified at any time during the
3455 /* request: hash to start with:
3456 * reply: hash of the first entry of the first lu_idxpage and hash
3457 * of the entry to read next if any */
3458 __u64 ii_hash_start
;
3461 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3465 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3473 extern void lustre_swab_idx_info(struct idx_info
*ii
);
3475 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3477 /* List of flags used in idx_info::ii_flags */
3478 enum idx_info_flags
{
3479 II_FL_NOHASH
= 1 << 0, /* client doesn't care about hash value */
3480 II_FL_VARKEY
= 1 << 1, /* keys can be of variable size */
3481 II_FL_VARREC
= 1 << 2, /* records can be of variable size */
3482 II_FL_NONUNQ
= 1 << 3, /* index supports non-unique keys */
3485 #define LIP_MAGIC 0x8A6D6B6C
3487 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3489 /* 16-byte header */
3492 __u16 lip_nr
; /* number of entries in the container */
3493 __u64 lip_pad0
; /* additional padding for future use */
3495 /* key/record pairs are stored in the remaining 4080 bytes.
3496 * depending upon the flags in idx_info::ii_flags, each key/record
3497 * pair might be preceded by:
3499 * - the key size (II_FL_VARKEY is set)
3500 * - the record size (II_FL_VARREC is set)
3502 * For the time being, we only support fixed-size key & record. */
3503 char lip_entries
[0];
3505 extern void lustre_swab_lip_header(struct lu_idxpage
*lip
);
3507 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3509 /* Gather all possible type associated with a 4KB container */
3511 struct lu_dirpage lp_dir
; /* for MDS_READPAGE */
3512 struct lu_idxpage lp_idx
; /* for OBD_IDX_READ */
3513 char lp_array
[LU_PAGE_SIZE
];
3516 /* security opcodes */
3519 SEC_CTX_INIT_CONT
= 802,
3522 SEC_FIRST_OPC
= SEC_CTX_INIT
3526 * capa related definitions
3528 #define CAPA_HMAC_MAX_LEN 64
3529 #define CAPA_HMAC_KEY_MAX_LEN 56
3531 /* NB take care when changing the sequence of elements this struct,
3532 * because the offset info is used in find_capa() */
3533 struct lustre_capa
{
3534 struct lu_fid lc_fid
; /** fid */
3535 __u64 lc_opc
; /** operations allowed */
3536 __u64 lc_uid
; /** file owner */
3537 __u64 lc_gid
; /** file group */
3538 __u32 lc_flags
; /** HMAC algorithm & flags */
3539 __u32 lc_keyid
; /** key# used for the capability */
3540 __u32 lc_timeout
; /** capa timeout value (sec) */
3541 __u32 lc_expiry
; /** expiry time (sec) */
3542 __u8 lc_hmac
[CAPA_HMAC_MAX_LEN
]; /** HMAC */
3543 } __attribute__((packed
));
3545 extern void lustre_swab_lustre_capa(struct lustre_capa
*c
);
3547 /** lustre_capa::lc_opc */
3549 CAPA_OPC_BODY_WRITE
= 1<<0, /**< write object data */
3550 CAPA_OPC_BODY_READ
= 1<<1, /**< read object data */
3551 CAPA_OPC_INDEX_LOOKUP
= 1<<2, /**< lookup object fid */
3552 CAPA_OPC_INDEX_INSERT
= 1<<3, /**< insert object fid */
3553 CAPA_OPC_INDEX_DELETE
= 1<<4, /**< delete object fid */
3554 CAPA_OPC_OSS_WRITE
= 1<<5, /**< write oss object data */
3555 CAPA_OPC_OSS_READ
= 1<<6, /**< read oss object data */
3556 CAPA_OPC_OSS_TRUNC
= 1<<7, /**< truncate oss object */
3557 CAPA_OPC_OSS_DESTROY
= 1<<8, /**< destroy oss object */
3558 CAPA_OPC_META_WRITE
= 1<<9, /**< write object meta data */
3559 CAPA_OPC_META_READ
= 1<<10, /**< read object meta data */
3562 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3563 #define CAPA_OPC_MDS_ONLY \
3564 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3565 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3566 #define CAPA_OPC_OSS_ONLY \
3567 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3568 CAPA_OPC_OSS_DESTROY)
3569 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3570 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3572 /* MDS capability covers object capability for operations of body r/w
3573 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
3574 * while OSS capability only covers object capability for operations of
3575 * oss data(file content) r/w/truncate.
3577 static inline int capa_for_mds(struct lustre_capa
*c
)
3579 return (c
->lc_opc
& CAPA_OPC_INDEX_LOOKUP
) != 0;
3582 static inline int capa_for_oss(struct lustre_capa
*c
)
3584 return (c
->lc_opc
& CAPA_OPC_INDEX_LOOKUP
) == 0;
3587 /* lustre_capa::lc_hmac_alg */
3589 CAPA_HMAC_ALG_SHA1
= 1, /**< sha1 algorithm */
3593 #define CAPA_FL_MASK 0x00ffffff
3594 #define CAPA_HMAC_ALG_MASK 0xff000000
3596 struct lustre_capa_key
{
3597 __u64 lk_seq
; /**< mds# */
3598 __u32 lk_keyid
; /**< key# */
3600 __u8 lk_key
[CAPA_HMAC_KEY_MAX_LEN
]; /**< key */
3601 } __attribute__((packed
));
3603 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key
*k
);
3605 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3606 #define LINK_EA_MAGIC 0x11EAF1DFUL
3607 struct link_ea_header
{
3610 __u64 leh_len
; /* total size */
3616 /** Hardlink data is name and parent fid.
3617 * Stored in this crazy struct for maximum packing and endian-neutrality
3619 struct link_ea_entry
{
3620 /** __u16 stored big-endian, unaligned */
3621 unsigned char lee_reclen
[2];
3622 unsigned char lee_parent_fid
[sizeof(struct lu_fid
)];
3624 }__attribute__((packed
));
3626 /** fid2path request/reply structure */
3627 struct getinfo_fid2path
{
3628 struct lu_fid gf_fid
;
3633 } __attribute__((packed
));
3635 void lustre_swab_fid2path (struct getinfo_fid2path
*gf
);
3638 LAYOUT_INTENT_ACCESS
= 0,
3639 LAYOUT_INTENT_READ
= 1,
3640 LAYOUT_INTENT_WRITE
= 2,
3641 LAYOUT_INTENT_GLIMPSE
= 3,
3642 LAYOUT_INTENT_TRUNC
= 4,
3643 LAYOUT_INTENT_RELEASE
= 5,
3644 LAYOUT_INTENT_RESTORE
= 6
3647 /* enqueue layout lock with intent */
3648 struct layout_intent
{
3649 __u32 li_opc
; /* intent operation for enqueue, read, write etc */
3655 void lustre_swab_layout_intent(struct layout_intent
*li
);
3658 * On the wire version of hsm_progress structure.
3660 * Contains the userspace hsm_progress and some internal fields.
3662 struct hsm_progress_kernel
{
3663 /* Field taken from struct hsm_progress */
3666 struct hsm_extent hpk_extent
;
3668 __u16 hpk_errval
; /* positive val */
3670 /* Additional fields */
3671 __u64 hpk_data_version
;
3673 } __attribute__((packed
));
3675 extern void lustre_swab_hsm_user_state(struct hsm_user_state
*hus
);
3676 extern void lustre_swab_hsm_current_action(struct hsm_current_action
*action
);
3677 extern void lustre_swab_hsm_progress_kernel(struct hsm_progress_kernel
*hpk
);
3678 extern void lustre_swab_hsm_user_state(struct hsm_user_state
*hus
);
3679 extern void lustre_swab_hsm_user_item(struct hsm_user_item
*hui
);
3680 extern void lustre_swab_hsm_request(struct hsm_request
*hr
);
3683 * These are object update opcode under UPDATE_OBJ, which is currently
3684 * being used by cross-ref operations between MDT.
3686 * During the cross-ref operation, the Master MDT, which the client send the
3687 * request to, will disassembly the operation into object updates, then OSP
3688 * will send these updates to the remote MDT to be executed.
3690 * Update request format
3691 * magic: UPDATE_BUFFER_MAGIC_V1
3692 * Count: How many updates in the req.
3693 * bufs[0] : following are packets of object.
3695 * type: object_update_op, the op code of update
3696 * fid: The object fid of the update.
3697 * lens/bufs: other parameters of the update.
3699 * type: object_update_op, the op code of update
3700 * fid: The object fid of the update.
3701 * lens/bufs: other parameters of the update.
3703 * update[7]: type: object_update_op, the op code of update
3704 * fid: The object fid of the update.
3705 * lens/bufs: other parameters of the update.
3706 * Current 8 maxim updates per object update request.
3708 *******************************************************************
3709 * update reply format:
3711 * ur_version: UPDATE_REPLY_V1
3712 * ur_count: The count of the reply, which is usually equal
3713 * to the number of updates in the request.
3714 * ur_lens: The reply lengths of each object update.
3716 * replies: 1st update reply [4bytes_ret: other body]
3717 * 2nd update reply [4bytes_ret: other body]
3719 * nth update reply [4bytes_ret: other body]
3721 * For each reply of the update, the format would be
3722 * result(4 bytes):Other stuff
3725 #define UPDATE_MAX_OPS 10
3726 #define UPDATE_BUFFER_MAGIC_V1 0xBDDE0001
3727 #define UPDATE_BUFFER_MAGIC UPDATE_BUFFER_MAGIC_V1
3728 #define UPDATE_BUF_COUNT 8
3729 enum object_update_op
{
3738 OBJ_INDEX_LOOKUP
= 9,
3739 OBJ_INDEX_INSERT
= 10,
3740 OBJ_INDEX_DELETE
= 11,
3747 struct lu_fid u_fid
;
3748 __u32 u_lens
[UPDATE_BUF_COUNT
];
3758 #define UPDATE_REPLY_V1 0x00BD0001
3759 struct update_reply
{
3765 void lustre_swab_update_buf(struct update_buf
*ub
);
3766 void lustre_swab_update_reply_buf(struct update_reply
*ur
);
3768 /** layout swap request structure
3769 * fid1 and fid2 are in mdt_body
3771 struct mdc_swap_layouts
{
3775 void lustre_swab_swap_layouts(struct mdc_swap_layouts
*msl
);
3778 struct lustre_handle cd_handle
;
3779 struct lu_fid cd_fid
;
3780 __u64 cd_data_version
;
3781 __u64 cd_reserved
[8];
3784 void lustre_swab_close_data(struct close_data
*data
);