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 #if !defined(LASSERT) && !defined(LPU64)
95 #include <libcfs/libcfs.h> /* for LASSERT, LPUX64, etc */
98 /* Defn's shared with user-space. */
99 #include <lustre/lustre_user.h>
104 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
105 * FOO_REPLY_PORTAL is for incoming replies on the FOO
106 * FOO_BULK_PORTAL is for incoming bulk on the FOO
109 #define CONNMGR_REQUEST_PORTAL 1
110 #define CONNMGR_REPLY_PORTAL 2
111 //#define OSC_REQUEST_PORTAL 3
112 #define OSC_REPLY_PORTAL 4
113 //#define OSC_BULK_PORTAL 5
114 #define OST_IO_PORTAL 6
115 #define OST_CREATE_PORTAL 7
116 #define OST_BULK_PORTAL 8
117 //#define MDC_REQUEST_PORTAL 9
118 #define MDC_REPLY_PORTAL 10
119 //#define MDC_BULK_PORTAL 11
120 #define MDS_REQUEST_PORTAL 12
121 //#define MDS_REPLY_PORTAL 13
122 #define MDS_BULK_PORTAL 14
123 #define LDLM_CB_REQUEST_PORTAL 15
124 #define LDLM_CB_REPLY_PORTAL 16
125 #define LDLM_CANCEL_REQUEST_PORTAL 17
126 #define LDLM_CANCEL_REPLY_PORTAL 18
127 //#define PTLBD_REQUEST_PORTAL 19
128 //#define PTLBD_REPLY_PORTAL 20
129 //#define PTLBD_BULK_PORTAL 21
130 #define MDS_SETATTR_PORTAL 22
131 #define MDS_READPAGE_PORTAL 23
132 #define MDS_MDS_PORTAL 24
134 #define MGC_REPLY_PORTAL 25
135 #define MGS_REQUEST_PORTAL 26
136 #define MGS_REPLY_PORTAL 27
137 #define OST_REQUEST_PORTAL 28
138 #define FLD_REQUEST_PORTAL 29
139 #define SEQ_METADATA_PORTAL 30
140 #define SEQ_DATA_PORTAL 31
141 #define SEQ_CONTROLLER_PORTAL 32
142 #define MGS_BULK_PORTAL 33
144 /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com, n8851@cray.com */
147 #define PTL_RPC_MSG_REQUEST 4711
148 #define PTL_RPC_MSG_ERR 4712
149 #define PTL_RPC_MSG_REPLY 4713
151 /* DON'T use swabbed values of MAGIC as magic! */
152 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
153 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
155 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
156 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
158 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
160 #define PTLRPC_MSG_VERSION 0x00000003
161 #define LUSTRE_VERSION_MASK 0xffff0000
162 #define LUSTRE_OBD_VERSION 0x00010000
163 #define LUSTRE_MDS_VERSION 0x00020000
164 #define LUSTRE_OST_VERSION 0x00030000
165 #define LUSTRE_DLM_VERSION 0x00040000
166 #define LUSTRE_LOG_VERSION 0x00050000
167 #define LUSTRE_MGS_VERSION 0x00060000
169 typedef __u32 mdsno_t;
170 typedef __u64 seqno_t;
171 typedef __u64 obd_id;
172 typedef __u64 obd_seq;
173 typedef __s64 obd_time;
174 typedef __u64 obd_size;
175 typedef __u64 obd_off;
176 typedef __u64 obd_blocks;
177 typedef __u64 obd_valid;
178 typedef __u32 obd_blksize;
179 typedef __u32 obd_mode;
180 typedef __u32 obd_uid;
181 typedef __u32 obd_gid;
182 typedef __u32 obd_flag;
183 typedef __u32 obd_count;
186 * Describes a range of sequence, lsr_start is included but lsr_end is
188 * Same structure is used in fld module where lsr_index field holds mdt id
192 #define LU_SEQ_RANGE_MDT 0x0
193 #define LU_SEQ_RANGE_OST 0x1
195 struct lu_seq_range {
203 * returns width of given range \a r
206 static inline __u64 range_space(const struct lu_seq_range *range)
208 return range->lsr_end - range->lsr_start;
212 * initialize range to zero
215 static inline void range_init(struct lu_seq_range *range)
217 range->lsr_start = range->lsr_end = range->lsr_index = 0;
221 * check if given seq id \a s is within given range \a r
224 static inline int range_within(const struct lu_seq_range *range,
227 return s >= range->lsr_start && s < range->lsr_end;
230 static inline int range_is_sane(const struct lu_seq_range *range)
232 return (range->lsr_end >= range->lsr_start);
235 static inline int range_is_zero(const struct lu_seq_range *range)
237 return (range->lsr_start == 0 && range->lsr_end == 0);
240 static inline int range_is_exhausted(const struct lu_seq_range *range)
243 return range_space(range) == 0;
246 /* return 0 if two range have the same location */
247 static inline int range_compare_loc(const struct lu_seq_range *r1,
248 const struct lu_seq_range *r2)
250 return r1->lsr_index != r2->lsr_index ||
251 r1->lsr_flags != r2->lsr_flags;
254 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x):%x:%s"
256 #define PRANGE(range) \
257 (range)->lsr_start, \
259 (range)->lsr_index, \
260 (range)->lsr_flags == LU_SEQ_RANGE_MDT ? "mdt" : "ost"
263 /** \defgroup lu_fid lu_fid
267 * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat.
268 * Deprecated since HSM and SOM attributes are now stored in separate on-disk
272 LMAC_HSM = 0x00000001,
273 LMAC_SOM = 0x00000002,
277 * Masks for all features that should be supported by a Lustre version to
278 * access a specific file.
279 * This information is stored in lustre_mdt_attrs::lma_incompat.
282 LMAI_RELEASED = 0x0000001, /* file is released */
284 #define LMA_INCOMPAT_SUPP 0x0
287 * Following struct for MDT attributes, that will be kept inode's EA.
288 * Introduced in 2.0 release (please see b15993, for details)
290 struct lustre_mdt_attrs {
292 * Bitfield for supported data in this structure. From enum lma_compat.
293 * lma_self_fid and lma_flags are always available.
297 * Per-file incompat feature list. Lustre version should support all
298 * flags set in this field. The supported feature mask is available in
302 /** FID of this inode */
303 struct lu_fid lma_self_fid;
304 /** mdt/ost type, others */
309 * Prior to 2.4, the LMA structure also included SOM attributes which has since
310 * been moved to a dedicated xattr
312 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 4 * sizeof(__u64))
314 extern void lustre_lma_swab(struct lustre_mdt_attrs *lma);
315 extern void lustre_lma_init(struct lustre_mdt_attrs *lma,
316 const struct lu_fid *fid);
318 * SOM on-disk attributes stored in a separate xattr.
321 /** Bitfield for supported data in this structure. For future use. */
324 /** Incompat feature list. The supported feature mask is availabe in
325 * SOM_INCOMPAT_SUPP */
328 /** IO Epoch SOM attributes belongs to */
330 /** total file size in objects */
332 /** total fs blocks in objects */
334 /** mds mount id the size is valid for */
337 extern void lustre_som_swab(struct som_attrs *attrs);
339 #define SOM_INCOMPAT_SUPP 0x0
342 * HSM on-disk attributes stored in a separate xattr.
345 /** Bitfield for supported data in this structure. For future use. */
348 /** HSM flags, see hsm_flags enum below */
350 /** backend archive id associated with the file */
352 /** version associated with the last archiving, if any */
355 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
361 /** initial fid id value */
362 LUSTRE_FID_INIT_OID = 1UL
365 /** returns fid object sequence */
366 static inline __u64 fid_seq(const struct lu_fid *fid)
371 /** returns fid object id */
372 static inline __u32 fid_oid(const struct lu_fid *fid)
377 /** returns fid object version */
378 static inline __u32 fid_ver(const struct lu_fid *fid)
383 static inline void fid_zero(struct lu_fid *fid)
385 memset(fid, 0, sizeof(*fid));
388 static inline obd_id fid_ver_oid(const struct lu_fid *fid)
390 return ((__u64)fid_ver(fid) << 32 | fid_oid(fid));
394 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
395 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
396 * used for other purposes and not risk collisions with existing inodes.
398 * Different FID Format
399 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
402 FID_SEQ_OST_MDT0 = 0,
405 FID_SEQ_OST_MDT1 = 3,
406 FID_SEQ_OST_MAX = 9, /* Max MDT count before OST_on_FID */
409 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
410 FID_SEQ_IDIF = 0x100000000ULL,
411 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
412 /* Normal FID sequence starts from this value, i.e. 1<<33 */
413 FID_SEQ_START = 0x200000000ULL,
414 /* sequence for local pre-defined FIDs listed in local_oid */
415 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
416 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
417 /* sequence is used for local named objects FIDs generated
418 * by local_object_storage library */
419 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
420 FID_SEQ_SPECIAL = 0x200000004ULL,
421 FID_SEQ_QUOTA = 0x200000005ULL,
422 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
423 FID_SEQ_NORMAL = 0x200000400ULL,
424 FID_SEQ_LOV_DEFAULT= 0xffffffffffffffffULL
427 #define OBIF_OID_MAX_BITS 32
428 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
429 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
430 #define IDIF_OID_MAX_BITS 48
431 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
432 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
434 /** OID for FID_SEQ_SPECIAL */
436 /* Big Filesystem Lock to serialize rename operations */
437 FID_OID_SPECIAL_BFL = 1UL,
440 /** OID for FID_SEQ_DOT_LUSTRE */
441 enum dot_lustre_oid {
442 FID_OID_DOT_LUSTRE = 1UL,
443 FID_OID_DOT_LUSTRE_OBF = 2UL,
446 static inline int fid_seq_is_mdt0(obd_seq seq)
448 return (seq == FID_SEQ_OST_MDT0);
451 static inline int fid_seq_is_mdt(const __u64 seq)
453 return seq == FID_SEQ_OST_MDT0 || seq >= FID_SEQ_NORMAL;
456 static inline int fid_seq_is_echo(obd_seq seq)
458 return (seq == FID_SEQ_ECHO);
461 static inline int fid_is_echo(const struct lu_fid *fid)
463 return fid_seq_is_echo(fid_seq(fid));
466 static inline int fid_seq_is_llog(obd_seq seq)
468 return (seq == FID_SEQ_LLOG);
471 static inline int fid_is_llog(const struct lu_fid *fid)
473 return fid_seq_is_llog(fid_seq(fid));
476 static inline int fid_seq_is_rsvd(const __u64 seq)
478 return (seq > FID_SEQ_OST_MDT0 && seq <= FID_SEQ_RSVD);
481 static inline int fid_is_mdt0(const struct lu_fid *fid)
483 return fid_seq_is_mdt0(fid_seq(fid));
487 * Check if a fid is igif or not.
488 * \param fid the fid to be tested.
489 * \return true if the fid is a igif; otherwise false.
491 static inline int fid_seq_is_igif(const __u64 seq)
493 return seq >= FID_SEQ_IGIF && seq <= FID_SEQ_IGIF_MAX;
496 static inline int fid_is_igif(const struct lu_fid *fid)
498 return fid_seq_is_igif(fid_seq(fid));
502 * Check if a fid is idif or not.
503 * \param fid the fid to be tested.
504 * \return true if the fid is a idif; otherwise false.
506 static inline int fid_seq_is_idif(const __u64 seq)
508 return seq >= FID_SEQ_IDIF && seq <= FID_SEQ_IDIF_MAX;
511 static inline int fid_is_idif(const struct lu_fid *fid)
513 return fid_seq_is_idif(fid_seq(fid));
521 static inline int fid_seq_is_norm(const __u64 seq)
523 return (seq >= FID_SEQ_NORMAL);
526 static inline int fid_is_norm(const struct lu_fid *fid)
528 return fid_seq_is_norm(fid_seq(fid));
531 /* convert an OST objid into an IDIF FID SEQ number */
532 static inline obd_seq fid_idif_seq(obd_id id, __u32 ost_idx)
534 return FID_SEQ_IDIF | (ost_idx << 16) | ((id >> 32) & 0xffff);
537 /* convert a packed IDIF FID into an OST objid */
538 static inline obd_id fid_idif_id(obd_seq seq, __u32 oid, __u32 ver)
540 return ((__u64)ver << 48) | ((seq & 0xffff) << 32) | oid;
543 /* extract ost index from IDIF FID */
544 static inline __u32 fid_idif_ost_idx(const struct lu_fid *fid)
546 LASSERT(fid_is_idif(fid));
547 return (fid_seq(fid) >> 16) & 0xffff;
550 /* unpack an ostid (id/seq) from a wire/disk structure into an IDIF FID */
551 static inline void ostid_idif_unpack(struct ost_id *ostid,
552 struct lu_fid *fid, __u32 ost_idx)
554 fid->f_seq = fid_idif_seq(ostid->oi_id, ost_idx);
555 fid->f_oid = ostid->oi_id; /* truncate to 32 bits by assignment */
556 fid->f_ver = ostid->oi_id >> 48; /* in theory, not currently used */
559 /* unpack an ostid (id/seq) from a wire/disk structure into a non-IDIF FID */
560 static inline void ostid_fid_unpack(struct ost_id *ostid, struct lu_fid *fid)
562 fid->f_seq = ostid->oi_seq;
563 fid->f_oid = ostid->oi_id; /* truncate to 32 bits by assignment */
564 fid->f_ver = ostid->oi_id >> 32; /* in theory, not currently used */
567 /* Unpack an OST object id/seq (group) into a FID. This is needed for
568 * converting all obdo, lmm, lsm, etc. 64-bit id/seq pairs into proper
569 * FIDs. Note that if an id/seq is already in FID/IDIF format it will
570 * be passed through unchanged. Only legacy OST objects in "group 0"
571 * will be mapped into the IDIF namespace so that they can fit into the
572 * struct lu_fid fields without loss. For reference see:
573 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs
575 static inline int fid_ostid_unpack(struct lu_fid *fid, struct ost_id *ostid,
578 if (ost_idx > 0xffff) {
579 CERROR("bad ost_idx, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
580 ostid->oi_seq, ostid->oi_id, ost_idx);
584 if (fid_seq_is_mdt0(ostid->oi_seq)) {
585 /* This is a "legacy" (old 1.x/2.early) OST object in "group 0"
586 * that we map into the IDIF namespace. It allows up to 2^48
587 * objects per OST, as this is the object namespace that has
588 * been in production for years. This can handle create rates
589 * of 1M objects/s/OST for 9 years, or combinations thereof. */
590 if (ostid->oi_id >= IDIF_MAX_OID) {
591 CERROR("bad MDT0 id, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
592 ostid->oi_seq, ostid->oi_id, ost_idx);
595 ostid_idif_unpack(ostid, fid, ost_idx);
597 } else if (fid_seq_is_rsvd(ostid->oi_seq)) {
598 /* These are legacy OST objects for LLOG/ECHO and CMD testing.
599 * We only support 2^32 objects in these groups, and cannot
600 * uniquely identify them in the system (i.e. they are the
601 * duplicated on all OSTs), but this is not strictly required
602 * for the old object protocol, which has a separate ost_idx. */
603 if (ostid->oi_id >= 0xffffffffULL) {
604 CERROR("bad RSVD id, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
605 ostid->oi_seq, ostid->oi_id, ost_idx);
608 ostid_fid_unpack(ostid, fid);
610 } else if (unlikely(fid_seq_is_igif(ostid->oi_seq))) {
611 /* This is an MDT inode number, which should never collide with
612 * proper OST object IDs, and is probably a broken filesystem */
613 CERROR("bad IGIF, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
614 ostid->oi_seq, ostid->oi_id, ost_idx);
617 } else /* if (fid_seq_is_idif(seq) || fid_seq_is_norm(seq)) */ {
618 /* This is either an IDIF object, which identifies objects across
619 * all OSTs, or a regular FID. The IDIF namespace maps legacy
620 * OST objects into the FID namespace. In both cases, we just
621 * pass the FID through, no conversion needed. */
622 ostid_fid_unpack(ostid, fid);
628 /* pack an IDIF FID into an ostid (id/seq) for the wire/disk */
629 static inline void ostid_idif_pack(const struct lu_fid *fid,
630 struct ost_id *ostid)
632 ostid->oi_seq = FID_SEQ_OST_MDT0;
633 ostid->oi_id = fid_idif_id(fid->f_seq, fid->f_oid, fid->f_ver);
636 /* pack a non-IDIF FID into an ostid (id/seq) for the wire/disk */
637 static inline void ostid_fid_pack(const struct lu_fid *fid,
638 struct ost_id *ostid)
640 ostid->oi_seq = fid_seq(fid);
641 ostid->oi_id = fid_ver_oid(fid);
644 /* pack any OST FID into an ostid (id/seq) for the wire/disk */
645 static inline int fid_ostid_pack(const struct lu_fid *fid,
646 struct ost_id *ostid)
648 if (unlikely(fid_seq_is_igif(fid->f_seq))) {
649 CERROR("bad IGIF, "DFID"\n", PFID(fid));
653 if (fid_is_idif(fid))
654 ostid_idif_pack(fid, ostid);
656 ostid_fid_pack(fid, ostid);
661 /* extract OST sequence (group) from a wire ost_id (id/seq) pair */
662 static inline obd_seq ostid_seq(struct ost_id *ostid)
664 if (unlikely(fid_seq_is_igif(ostid->oi_seq)))
665 CWARN("bad IGIF, oi_seq: "LPU64" oi_id: "LPX64"\n",
666 ostid->oi_seq, ostid->oi_id);
668 if (unlikely(fid_seq_is_idif(ostid->oi_seq)))
669 return FID_SEQ_OST_MDT0;
671 return ostid->oi_seq;
674 /* extract OST objid from a wire ost_id (id/seq) pair */
675 static inline obd_id ostid_id(struct ost_id *ostid)
677 if (ostid->oi_seq == FID_SEQ_OST_MDT0)
678 return ostid->oi_id & IDIF_OID_MASK;
680 if (fid_seq_is_rsvd(ostid->oi_seq))
681 return ostid->oi_id & OBIF_OID_MASK;
683 if (fid_seq_is_idif(ostid->oi_seq))
684 return fid_idif_id(ostid->oi_seq, ostid->oi_id, 0);
689 /* Check whether the fid is for LAST_ID */
690 static inline int fid_is_last_id(const struct lu_fid *fid)
692 return (fid_is_idif(fid) || fid_is_norm(fid) || fid_is_echo(fid)) &&
697 * Get inode number from a igif.
698 * \param fid a igif to get inode number from.
699 * \return inode number for the igif.
701 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
707 * Build igif from the inode number/generation.
709 #define LU_IGIF_BUILD(fid, ino, gen) \
715 static inline void lu_igif_build(struct lu_fid *fid, __u32 ino, __u32 gen)
717 LU_IGIF_BUILD(fid, ino, gen);
718 LASSERT(fid_is_igif(fid));
722 * Get inode generation from a igif.
723 * \param fid a igif to get inode generation from.
724 * \return inode generation for the igif.
726 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
732 * Fids are transmitted across network (in the sender byte-ordering),
733 * and stored on disk in big-endian order.
735 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
737 /* check that all fields are converted */
738 CLASSERT(sizeof *src ==
739 sizeof fid_seq(src) +
740 sizeof fid_oid(src) + sizeof fid_ver(src));
741 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
742 dst->f_seq = cpu_to_le64(fid_seq(src));
743 dst->f_oid = cpu_to_le32(fid_oid(src));
744 dst->f_ver = cpu_to_le32(fid_ver(src));
747 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
749 /* check that all fields are converted */
750 CLASSERT(sizeof *src ==
751 sizeof fid_seq(src) +
752 sizeof fid_oid(src) + sizeof fid_ver(src));
753 dst->f_seq = le64_to_cpu(fid_seq(src));
754 dst->f_oid = le32_to_cpu(fid_oid(src));
755 dst->f_ver = le32_to_cpu(fid_ver(src));
756 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
759 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
761 /* check that all fields are converted */
762 CLASSERT(sizeof *src ==
763 sizeof fid_seq(src) +
764 sizeof fid_oid(src) + sizeof fid_ver(src));
765 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
766 dst->f_seq = cpu_to_be64(fid_seq(src));
767 dst->f_oid = cpu_to_be32(fid_oid(src));
768 dst->f_ver = cpu_to_be32(fid_ver(src));
771 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
773 /* check that all fields are converted */
774 CLASSERT(sizeof *src ==
775 sizeof fid_seq(src) +
776 sizeof fid_oid(src) + sizeof fid_ver(src));
777 dst->f_seq = be64_to_cpu(fid_seq(src));
778 dst->f_oid = be32_to_cpu(fid_oid(src));
779 dst->f_ver = be32_to_cpu(fid_ver(src));
780 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
783 static inline int fid_is_sane(const struct lu_fid *fid)
785 return fid != NULL &&
786 ((fid_seq(fid) >= FID_SEQ_START && fid_ver(fid) == 0) ||
787 fid_is_igif(fid) || fid_is_idif(fid) ||
788 fid_seq_is_rsvd(fid_seq(fid)));
791 static inline int fid_is_zero(const struct lu_fid *fid)
793 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
796 extern void lustre_swab_lu_fid(struct lu_fid *fid);
797 extern void lustre_swab_lu_seq_range(struct lu_seq_range *range);
799 static inline int lu_fid_eq(const struct lu_fid *f0,
800 const struct lu_fid *f1)
802 /* Check that there is no alignment padding. */
803 CLASSERT(sizeof *f0 ==
804 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
805 LASSERTF((fid_is_igif(f0) || fid_is_idif(f0)) ||
806 fid_ver(f0) == 0, DFID, PFID(f0));
807 LASSERTF((fid_is_igif(f1) || fid_is_idif(f1)) ||
808 fid_ver(f1) == 0, DFID, PFID(f1));
809 return memcmp(f0, f1, sizeof *f0) == 0;
812 #define __diff_normalize(val0, val1) \
814 typeof(val0) __val0 = (val0); \
815 typeof(val1) __val1 = (val1); \
817 (__val0 == __val1 ? 0 : __val0 > __val1 ? +1 : -1); \
820 static inline int lu_fid_cmp(const struct lu_fid *f0,
821 const struct lu_fid *f1)
824 __diff_normalize(fid_seq(f0), fid_seq(f1)) ?:
825 __diff_normalize(fid_oid(f0), fid_oid(f1)) ?:
826 __diff_normalize(fid_ver(f0), fid_ver(f1));
831 /** \defgroup lu_dir lu_dir
835 * Enumeration of possible directory entry attributes.
837 * Attributes follow directory entry header in the order they appear in this
840 enum lu_dirent_attrs {
843 LUDA_64BITHASH = 0x0004,
847 * Layout of readdir pages, as transmitted on wire.
850 /** valid if LUDA_FID is set. */
851 struct lu_fid lde_fid;
852 /** a unique entry identifier: a hash or an offset. */
854 /** total record length, including all attributes. */
858 /** optional variable size attributes following this entry.
859 * taken from enum lu_dirent_attrs.
862 /** name is followed by the attributes indicated in ->ldp_attrs, in
863 * their natural order. After the last attribute, padding bytes are
864 * added to make ->lde_reclen a multiple of 8.
870 * Definitions of optional directory entry attributes formats.
872 * Individual attributes do not have their length encoded in a generic way. It
873 * is assumed that consumer of an attribute knows its format. This means that
874 * it is impossible to skip over an unknown attribute, except by skipping over all
875 * remaining attributes (by using ->lde_reclen), which is not too
876 * constraining, because new server versions will append new attributes at
877 * the end of an entry.
881 * Fid directory attribute: a fid of an object referenced by the entry. This
882 * will be almost always requested by the client and supplied by the server.
884 * Aligned to 8 bytes.
886 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
891 * Aligned to 2 bytes.
898 __u64 ldp_hash_start;
902 struct lu_dirent ldp_entries[0];
905 enum lu_dirpage_flags {
907 * dirpage contains no entry.
911 * last entry's lde_hash equals ldp_hash_end.
916 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
918 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
921 return dp->ldp_entries;
924 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
926 struct lu_dirent *next;
928 if (le16_to_cpu(ent->lde_reclen) != 0)
929 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
936 static inline int lu_dirent_calc_size(int namelen, __u16 attr)
940 if (attr & LUDA_TYPE) {
941 const unsigned align = sizeof(struct luda_type) - 1;
942 size = (sizeof(struct lu_dirent) + namelen + align) & ~align;
943 size += sizeof(struct luda_type);
945 size = sizeof(struct lu_dirent) + namelen;
947 return (size + 7) & ~7;
950 static inline int lu_dirent_size(struct lu_dirent *ent)
952 if (le16_to_cpu(ent->lde_reclen) == 0) {
953 return lu_dirent_calc_size(le16_to_cpu(ent->lde_namelen),
954 le32_to_cpu(ent->lde_attrs));
956 return le16_to_cpu(ent->lde_reclen);
959 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
962 * MDS_READPAGE page size
964 * This is the directory page size packed in MDS_READPAGE RPC.
965 * It's different than CFS_PAGE_SIZE because the client needs to
966 * access the struct lu_dirpage header packed at the beginning of
967 * the "page" and without this there isn't any way to know find the
968 * lu_dirpage header is if client and server CFS_PAGE_SIZE differ.
970 #define LU_PAGE_SHIFT 12
971 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
972 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
974 #define LU_PAGE_COUNT (1 << (CFS_PAGE_SHIFT - LU_PAGE_SHIFT))
978 struct lustre_handle {
981 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
983 static inline int lustre_handle_is_used(struct lustre_handle *lh)
985 return lh->cookie != 0ull;
988 static inline int lustre_handle_equal(const struct lustre_handle *lh1,
989 const struct lustre_handle *lh2)
991 return lh1->cookie == lh2->cookie;
994 static inline void lustre_handle_copy(struct lustre_handle *tgt,
995 struct lustre_handle *src)
997 tgt->cookie = src->cookie;
1000 /* flags for lm_flags */
1001 #define MSGHDR_AT_SUPPORT 0x1
1002 #define MSGHDR_CKSUM_INCOMPAT18 0x2
1004 #define lustre_msg lustre_msg_v2
1005 /* we depend on this structure to be 8-byte aligned */
1006 /* this type is only endian-adjusted in lustre_unpack_msg() */
1007 struct lustre_msg_v2 {
1016 __u32 lm_buflens[0];
1019 /* without gss, ptlrpc_body is put at the first buffer. */
1020 #define PTLRPC_NUM_VERSIONS 4
1021 #define JOBSTATS_JOBID_SIZE 32 /* 32 bytes string */
1022 struct ptlrpc_body_v3 {
1023 struct lustre_handle pb_handle;
1030 __u64 pb_last_committed;
1035 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1036 __u32 pb_service_time; /* for rep, actual service time */
1039 /* VBR: pre-versions */
1040 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1041 /* padding for future needs */
1042 __u64 pb_padding[4];
1043 char pb_jobid[JOBSTATS_JOBID_SIZE];
1045 #define ptlrpc_body ptlrpc_body_v3
1047 struct ptlrpc_body_v2 {
1048 struct lustre_handle pb_handle;
1055 __u64 pb_last_committed;
1060 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1061 __u32 pb_service_time; /* for rep, actual service time, also used for
1062 net_latency of req */
1065 /* VBR: pre-versions */
1066 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1067 /* padding for future needs */
1068 __u64 pb_padding[4];
1071 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
1073 /* message body offset for lustre_msg_v2 */
1074 /* ptlrpc body offset in all request/reply messages */
1075 #define MSG_PTLRPC_BODY_OFF 0
1077 /* normal request/reply message record offset */
1078 #define REQ_REC_OFF 1
1079 #define REPLY_REC_OFF 1
1081 /* ldlm request message body offset */
1082 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
1083 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
1085 /* ldlm intent lock message body offset */
1086 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
1087 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
1089 /* ldlm reply message body offset */
1090 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
1091 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
1093 /** only use in req->rq_{req,rep}_swab_mask */
1094 #define MSG_PTLRPC_HEADER_OFF 31
1096 /* Flags that are operation-specific go in the top 16 bits. */
1097 #define MSG_OP_FLAG_MASK 0xffff0000
1098 #define MSG_OP_FLAG_SHIFT 16
1100 /* Flags that apply to all requests are in the bottom 16 bits */
1101 #define MSG_GEN_FLAG_MASK 0x0000ffff
1102 #define MSG_LAST_REPLAY 0x0001
1103 #define MSG_RESENT 0x0002
1104 #define MSG_REPLAY 0x0004
1105 /* #define MSG_AT_SUPPORT 0x0008
1106 * This was used in early prototypes of adaptive timeouts, and while there
1107 * shouldn't be any users of that code there also isn't a need for using this
1108 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
1109 #define MSG_DELAY_REPLAY 0x0010
1110 #define MSG_VERSION_REPLAY 0x0020
1111 #define MSG_REQ_REPLAY_DONE 0x0040
1112 #define MSG_LOCK_REPLAY_DONE 0x0080
1115 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
1118 #define MSG_CONNECT_RECOVERING 0x00000001
1119 #define MSG_CONNECT_RECONNECT 0x00000002
1120 #define MSG_CONNECT_REPLAYABLE 0x00000004
1121 //#define MSG_CONNECT_PEER 0x8
1122 #define MSG_CONNECT_LIBCLIENT 0x00000010
1123 #define MSG_CONNECT_INITIAL 0x00000020
1124 #define MSG_CONNECT_ASYNC 0x00000040
1125 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
1126 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
1129 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
1130 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
1131 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
1132 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
1133 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
1134 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
1135 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
1136 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
1137 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
1138 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
1139 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
1140 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
1141 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
1142 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
1143 *We do not support JOIN FILE
1144 *anymore, reserve this flags
1145 *just for preventing such bit
1147 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
1148 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
1149 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /*Remote client */
1150 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /*Remote client by force */
1151 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
1152 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
1153 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
1154 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
1155 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
1156 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
1157 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
1158 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
1159 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
1160 #define OBD_CONNECT_REAL 0x8000000ULL /*real connection */
1161 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
1162 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
1163 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
1164 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
1165 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
1166 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
1167 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
1168 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
1169 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
1170 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
1171 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
1173 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
1174 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
1175 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
1176 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
1177 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
1178 * RPC error properly */
1179 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
1180 * finer space reservation */
1181 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
1182 * policy and 2.x server */
1183 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
1184 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
1185 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
1186 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
1188 * Please DO NOT add flag values here before first ensuring that this same
1189 * flag value is not in use on some other branch. Please clear any such
1190 * changes with senior engineers before starting to use a new flag. Then,
1191 * submit a small patch against EVERY branch that ONLY adds the new flag
1192 * and updates obd_connect_names[] for lprocfs_rd_connect_flags(), so it
1193 * can be approved and landed easily to reserve the flag for future use. */
1195 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
1196 * connection. It is a temporary bug fix for Imperative Recovery interop
1197 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
1198 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
1199 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
1201 #define OCD_HAS_FLAG(ocd, flg) \
1202 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
1205 #ifdef HAVE_LRU_RESIZE_SUPPORT
1206 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
1208 #define LRU_RESIZE_CONNECT_FLAG 0
1211 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
1212 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
1213 OBD_CONNECT_IBITS | \
1214 OBD_CONNECT_NODEVOH | OBD_CONNECT_ATTRFID | \
1215 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1216 OBD_CONNECT_RMT_CLIENT | \
1217 OBD_CONNECT_RMT_CLIENT_FORCE | \
1218 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_MDS_CAPA | \
1219 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_MDS_MDS | \
1220 OBD_CONNECT_FID | LRU_RESIZE_CONNECT_FLAG | \
1221 OBD_CONNECT_VBR | OBD_CONNECT_LOV_V3 | \
1222 OBD_CONNECT_SOM | OBD_CONNECT_FULL20 | \
1223 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
1224 OBD_CONNECT_EINPROGRESS | \
1225 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
1226 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK)
1227 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
1228 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
1229 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
1230 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_OSS_CAPA | \
1231 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1232 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_CKSUM | \
1233 OBD_CONNECT_RMT_CLIENT | \
1234 OBD_CONNECT_RMT_CLIENT_FORCE | OBD_CONNECT_VBR | \
1235 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
1236 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 | \
1237 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
1238 OBD_CONNECT_MAX_EASIZE | \
1239 OBD_CONNECT_EINPROGRESS | \
1240 OBD_CONNECT_JOBSTATS | \
1241 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
1242 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID)
1243 #define ECHO_CONNECT_SUPPORTED (0)
1244 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
1245 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
1246 OBD_CONNECT_MNE_SWAB)
1248 /* Features required for this version of the client to work with server */
1249 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
1252 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
1253 ((patch)<<8) + (fix))
1254 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
1255 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
1256 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
1257 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
1259 /* This structure is used for both request and reply.
1261 * If we eventually have separate connect data for different types, which we
1262 * almost certainly will, then perhaps we stick a union in here. */
1263 struct obd_connect_data_v1 {
1264 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
1265 __u32 ocd_version; /* lustre release version number */
1266 __u32 ocd_grant; /* initial cache grant amount (bytes) */
1267 __u32 ocd_index; /* LOV index to connect to */
1268 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
1269 __u64 ocd_ibits_known; /* inode bits this client understands */
1270 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
1271 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
1272 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
1273 __u32 ocd_unused; /* also fix lustre_swab_connect */
1274 __u64 ocd_transno; /* first transno from client to be replayed */
1275 __u32 ocd_group; /* MDS group on OST */
1276 __u32 ocd_cksum_types; /* supported checksum algorithms */
1277 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1278 __u32 ocd_instance; /* also fix lustre_swab_connect */
1279 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1282 struct obd_connect_data {
1283 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
1284 __u32 ocd_version; /* lustre release version number */
1285 __u32 ocd_grant; /* initial cache grant amount (bytes) */
1286 __u32 ocd_index; /* LOV index to connect to */
1287 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
1288 __u64 ocd_ibits_known; /* inode bits this client understands */
1289 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
1290 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
1291 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
1292 __u32 ocd_unused; /* also fix lustre_swab_connect */
1293 __u64 ocd_transno; /* first transno from client to be replayed */
1294 __u32 ocd_group; /* MDS group on OST */
1295 __u32 ocd_cksum_types; /* supported checksum algorithms */
1296 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1297 __u32 ocd_instance; /* instance # of this target */
1298 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1299 /* Fields after ocd_maxbytes are only accessible by the receiver
1300 * if the corresponding flag in ocd_connect_flags is set. Accessing
1301 * any field after ocd_maxbytes on the receiver without a valid flag
1302 * may result in out-of-bound memory access and kernel oops. */
1303 __u64 padding1; /* added 2.1.0. also fix lustre_swab_connect */
1304 __u64 padding2; /* added 2.1.0. also fix lustre_swab_connect */
1305 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
1306 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
1307 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
1308 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
1309 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
1310 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
1311 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
1312 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
1313 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
1314 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
1315 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
1316 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
1317 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1320 * Please DO NOT use any fields here before first ensuring that this same
1321 * field is not in use on some other branch. Please clear any such changes
1322 * with senior engineers before starting to use a new field. Then, submit
1323 * a small patch against EVERY branch that ONLY adds the new field along with
1324 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
1325 * reserve the flag for future use. */
1328 extern void lustre_swab_connect(struct obd_connect_data *ocd);
1331 * Supported checksum algorithms. Up to 32 checksum types are supported.
1332 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1333 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
1334 * algorithm and also the OBD_FL_CKSUM* flags.
1337 OBD_CKSUM_CRC32 = 0x00000001,
1338 OBD_CKSUM_ADLER = 0x00000002,
1339 OBD_CKSUM_CRC32C= 0x00000004,
1343 * OST requests: OBDO & OBD request records
1348 OST_REPLY = 0, /* reply ? */
1364 OST_QUOTACHECK = 18,
1366 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1369 #define OST_FIRST_OPC OST_REPLY
1372 OBD_FL_INLINEDATA = 0x00000001,
1373 OBD_FL_OBDMDEXISTS = 0x00000002,
1374 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1375 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1376 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1377 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1378 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1379 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1380 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1381 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1382 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1383 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1384 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1385 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1386 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
1387 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1388 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1389 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1390 * XXX: obsoleted - reserved for old
1391 * clients prior than 2.2 */
1392 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1393 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1395 /* Note that while these checksum values are currently separate bits,
1396 * in 2.x we can actually allow all values from 1-31 if we wanted. */
1397 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1398 OBD_FL_CKSUM_CRC32C,
1400 /* mask for local-only flag, which won't be sent over network */
1401 OBD_FL_LOCAL_MASK = 0xF0000000,
1404 #define LOV_MAGIC_V1 0x0BD10BD0
1405 #define LOV_MAGIC LOV_MAGIC_V1
1406 #define LOV_MAGIC_JOIN_V1 0x0BD20BD0
1407 #define LOV_MAGIC_V3 0x0BD30BD0
1410 * magic for fully defined striping
1411 * the idea is that we should have different magics for striping "hints"
1412 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1413 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1414 * we can't just change it w/o long way preparation, but we still need a
1415 * mechanism to allow LOD to differentiate hint versus ready striping.
1416 * so, at the moment we do a trick: MDT knows what to expect from request
1417 * depending on the case (replay uses ready striping, non-replay req uses
1418 * hints), so MDT replaces magic with appropriate one and now LOD can
1419 * easily understand what's inside -bzzz
1421 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
1422 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
1424 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
1425 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
1426 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
1427 #define LOV_PATTERN_CMOBD 0x200
1429 #define lov_ost_data lov_ost_data_v1
1430 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1431 __u64 l_object_id; /* OST object ID */
1432 __u64 l_object_seq; /* OST object seq number */
1433 __u32 l_ost_gen; /* generation of this l_ost_idx */
1434 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1437 #define lov_mds_md lov_mds_md_v1
1438 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1439 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1440 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1441 __u64 lmm_object_id; /* LOV object ID */
1442 __u64 lmm_object_seq; /* LOV object seq number */
1443 __u32 lmm_stripe_size; /* size of stripe in bytes */
1444 /* lmm_stripe_count used to be __u32 */
1445 __u16 lmm_stripe_count; /* num stripes in use for this object */
1446 __u16 lmm_layout_gen; /* layout generation number */
1447 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1450 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
1452 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1453 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1455 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1456 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1457 #define XATTR_USER_PREFIX "user."
1458 #define XATTR_TRUSTED_PREFIX "trusted."
1459 #define XATTR_SECURITY_PREFIX "security."
1460 #define XATTR_LUSTRE_PREFIX "lustre."
1462 #define XATTR_NAME_LOV "trusted.lov"
1463 #define XATTR_NAME_LMA "trusted.lma"
1464 #define XATTR_NAME_LMV "trusted.lmv"
1465 #define XATTR_NAME_LINK "trusted.link"
1466 #define XATTR_NAME_FID "trusted.fid"
1467 #define XATTR_NAME_VERSION "trusted.version"
1468 #define XATTR_NAME_SOM "trusted.som"
1469 #define XATTR_NAME_HSM "trusted.hsm"
1472 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1473 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1474 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1475 __u64 lmm_object_id; /* LOV object ID */
1476 __u64 lmm_object_seq; /* LOV object seq number */
1477 __u32 lmm_stripe_size; /* size of stripe in bytes */
1478 /* lmm_stripe_count used to be __u32 */
1479 __u16 lmm_stripe_count; /* num stripes in use for this object */
1480 __u16 lmm_layout_gen; /* layout generation number */
1481 char lmm_pool_name[LOV_MAXPOOLNAME]; /* must be 32bit aligned */
1482 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1486 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1487 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1488 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1489 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1490 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1491 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1492 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1493 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1494 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1495 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1496 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1497 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1498 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1499 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1500 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
1501 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1502 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1503 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1504 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1505 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1506 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1507 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1508 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
1509 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1510 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1511 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1512 /* ->mds if epoch opens or closes */
1513 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1514 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1515 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1516 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1517 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1519 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1520 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1521 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1523 /* OBD_MD_MDTIDX is used to get MDT index, but it is never been used overwire,
1524 * and it is already obsolete since 2.3 */
1525 /* #define OBD_MD_MDTIDX (0x0000000800000000ULL) */
1527 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1528 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1529 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1530 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1531 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
1532 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1533 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1534 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1535 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1536 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1538 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1540 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
1541 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
1542 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
1543 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
1545 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1547 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1548 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1549 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1550 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1551 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1553 /* don't forget obdo_fid which is way down at the bottom so it can
1554 * come after the definition of llog_cookie */
1558 HSS_CLEARMASK = 0x02,
1559 HSS_ARCHIVE_ID = 0x04,
1562 struct hsm_state_set {
1564 __u32 hss_archive_id;
1566 __u64 hss_clearmask;
1569 extern void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
1570 extern void lustre_swab_hsm_state_set(struct hsm_state_set *hss);
1572 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
1574 /* ost_body.data values for OST_BRW */
1576 #define OBD_BRW_READ 0x01
1577 #define OBD_BRW_WRITE 0x02
1578 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1579 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1580 * transfer and is not accounted in
1582 #define OBD_BRW_CHECK 0x10
1583 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1584 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1585 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1586 #define OBD_BRW_NOQUOTA 0x100
1587 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1588 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1589 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1590 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1591 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1593 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
1595 #define OST_MIN_PRECREATE 32
1596 #define OST_MAX_PRECREATE 20000
1599 struct ost_id ioo_oid;
1604 #define ioo_id ioo_oid.oi_id
1605 #define ioo_seq ioo_oid.oi_seq
1607 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
1609 /* multiple of 8 bytes => can array */
1610 struct niobuf_remote {
1616 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
1618 /* lock value block communicated between the filter and llite */
1620 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1621 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1622 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1623 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1624 #define OST_LVB_IS_ERR(blocks) \
1625 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1626 #define OST_LVB_SET_ERR(blocks, rc) \
1627 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1628 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1638 extern void lustre_swab_ost_lvb_v1(struct ost_lvb_v1 *lvb);
1652 extern void lustre_swab_ost_lvb(struct ost_lvb *lvb);
1655 * lquota data structures
1658 #ifndef QUOTABLOCK_BITS
1659 #define QUOTABLOCK_BITS 10
1662 #ifndef QUOTABLOCK_SIZE
1663 #define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS)
1667 #define toqb(x) (((x) + QUOTABLOCK_SIZE - 1) >> QUOTABLOCK_BITS)
1670 /* The lquota_id structure is an union of all the possible identifier types that
1671 * can be used with quota, this includes:
1674 * - a FID which can be used for per-directory quota in the future */
1676 struct lu_fid qid_fid; /* FID for per-directory quota */
1677 __u64 qid_uid; /* user identifier */
1678 __u64 qid_gid; /* group identifier */
1681 /* quotactl management */
1682 struct obd_quotactl {
1684 __u32 qc_type; /* see Q_* flag below */
1687 struct obd_dqinfo qc_dqinfo;
1688 struct obd_dqblk qc_dqblk;
1691 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1693 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1694 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1695 #define Q_GETOINFO 0x800102 /* get obd quota info */
1696 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1697 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1699 #define Q_COPY(out, in, member) (out)->member = (in)->member
1701 #define QCTL_COPY(out, in) \
1703 Q_COPY(out, in, qc_cmd); \
1704 Q_COPY(out, in, qc_type); \
1705 Q_COPY(out, in, qc_id); \
1706 Q_COPY(out, in, qc_stat); \
1707 Q_COPY(out, in, qc_dqinfo); \
1708 Q_COPY(out, in, qc_dqblk); \
1711 /* Body of quota request used for quota acquire/release RPCs between quota
1712 * master (aka QMT) and slaves (ak QSD). */
1714 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1715 * and type (data or metadata) as well as
1716 * the quota type (user or group). */
1717 union lquota_id qb_id; /* uid or gid or directory FID */
1718 __u32 qb_flags; /* see below */
1720 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1721 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1722 __u64 qb_slv_ver; /* slave index file version */
1723 struct lustre_handle qb_lockh; /* per-ID lock handle */
1724 struct lustre_handle qb_glb_lockh; /* global lock handle */
1725 __u64 qb_padding1[4];
1728 /* When the quota_body is used in the reply of quota global intent
1729 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1730 #define qb_slv_fid qb_fid
1731 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1733 #define qb_qunit qb_usage
1735 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1736 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1737 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1738 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1740 extern void lustre_swab_quota_body(struct quota_body *b);
1742 /* Quota types currently supported */
1744 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1745 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1749 /* There are 2 different resource types on which a quota limit can be enforced:
1750 * - inodes on the MDTs
1751 * - blocks on the OSTs */
1753 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1754 LQUOTA_RES_DT = 0x02,
1756 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1758 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1761 * Space accounting support
1762 * Format of an accounting record, providing disk usage information for a given
1765 struct lquota_acct_rec { /* 16 bytes */
1766 __u64 bspace; /* current space in use */
1767 __u64 ispace; /* current # inodes in use */
1771 * Global quota index support
1772 * Format of a global record, providing global quota settings for a given quota
1775 struct lquota_glb_rec { /* 32 bytes */
1776 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1777 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1778 __u64 qbr_time; /* grace time, in seconds */
1779 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1784 * Slave index support
1785 * Format of a slave record, recording how much space is granted to a given
1788 struct lquota_slv_rec { /* 8 bytes */
1789 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1790 * in #inodes or kbytes */
1793 /* Data structures associated with the quota locks */
1795 /* Glimpse descriptor used for the index & per-ID quota locks */
1796 struct ldlm_gl_lquota_desc {
1797 union lquota_id gl_id; /* quota ID subject to the glimpse */
1798 __u64 gl_flags; /* see LQUOTA_FL* below */
1799 __u64 gl_ver; /* new index version */
1800 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1801 __u64 gl_softlimit; /* new softlimit */
1805 #define gl_qunit gl_hardlimit /* current qunit value used when
1806 * glimpsing per-ID quota locks */
1808 /* quota glimpse flags */
1809 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1811 /* LVB used with quota (global and per-ID) locks */
1813 __u64 lvb_flags; /* see LQUOTA_FL* above */
1814 __u64 lvb_id_may_rel; /* space that might be released later */
1815 __u64 lvb_id_rel; /* space released by the slave for this ID */
1816 __u64 lvb_id_qunit; /* current qunit value */
1820 extern void lustre_swab_lquota_lvb(struct lquota_lvb *lvb);
1822 /* LVB used with global quota lock */
1823 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1831 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1840 MDS_GETATTR_NAME = 34,
1845 MDS_DISCONNECT = 39,
1851 MDS_DONE_WRITING = 45,
1853 MDS_QUOTACHECK = 47,
1856 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1860 MDS_HSM_STATE_GET = 54,
1861 MDS_HSM_STATE_SET = 55,
1862 MDS_HSM_ACTION = 56,
1863 MDS_HSM_PROGRESS = 57,
1864 MDS_HSM_REQUEST = 58,
1865 MDS_HSM_CT_REGISTER = 59,
1866 MDS_HSM_CT_UNREGISTER = 60,
1870 #define MDS_FIRST_OPC MDS_GETATTR
1887 } mds_reint_t, mdt_reint_t;
1889 extern void lustre_swab_generic_32s (__u32 *val);
1891 /* the disposition of the intent outlines what was executed */
1892 #define DISP_IT_EXECD 0x00000001
1893 #define DISP_LOOKUP_EXECD 0x00000002
1894 #define DISP_LOOKUP_NEG 0x00000004
1895 #define DISP_LOOKUP_POS 0x00000008
1896 #define DISP_OPEN_CREATE 0x00000010
1897 #define DISP_OPEN_OPEN 0x00000020
1898 #define DISP_ENQ_COMPLETE 0x00400000
1899 #define DISP_ENQ_OPEN_REF 0x00800000
1900 #define DISP_ENQ_CREATE_REF 0x01000000
1901 #define DISP_OPEN_LOCK 0x02000000
1903 /* INODE LOCK PARTS */
1904 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1905 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1906 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1907 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1909 #define MDS_INODELOCK_MAXSHIFT 3
1910 /* This FULL lock is useful to take on unlink sort of operations */
1911 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1913 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1915 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1916 * but was moved into name[1] along with the OID to avoid consuming the
1917 * name[2,3] fields that need to be used for the quota id (also a FID). */
1919 LUSTRE_RES_ID_SEQ_OFF = 0,
1920 LUSTRE_RES_ID_VER_OID_OFF = 1,
1921 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1922 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1923 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1924 LUSTRE_RES_ID_HSH_OFF = 3
1927 #define MDS_STATUS_CONN 1
1928 #define MDS_STATUS_LOV 2
1930 /* mdt_thread_info.mti_flags. */
1932 /* The flag indicates Size-on-MDS attributes are changed. */
1933 MF_SOM_CHANGE = (1 << 0),
1934 /* Flags indicates an epoch opens or closes. */
1935 MF_EPOCH_OPEN = (1 << 1),
1936 MF_EPOCH_CLOSE = (1 << 2),
1937 MF_MDC_CANCEL_FID1 = (1 << 3),
1938 MF_MDC_CANCEL_FID2 = (1 << 4),
1939 MF_MDC_CANCEL_FID3 = (1 << 5),
1940 MF_MDC_CANCEL_FID4 = (1 << 6),
1941 /* There is a pending attribute update. */
1942 MF_SOM_AU = (1 << 7),
1943 /* Cancel OST locks while getattr OST attributes. */
1944 MF_GETATTR_LOCK = (1 << 8),
1945 MF_GET_MDT_IDX = (1 << 9),
1948 #define MF_SOM_LOCAL_FLAGS (MF_SOM_CHANGE | MF_EPOCH_OPEN | MF_EPOCH_CLOSE)
1950 #define LUSTRE_BFLAG_UNCOMMITTED_WRITES 0x1
1952 /* these should be identical to their EXT4_*_FL counterparts, they are
1953 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1954 #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */
1955 #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */
1956 #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */
1957 #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */
1958 #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1961 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1962 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1963 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1964 * the S_* flags are kernel-internal values that change between kernel
1965 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1966 * See b=16526 for a full history. */
1967 static inline int ll_ext_to_inode_flags(int flags)
1969 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1970 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1971 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1972 #if defined(S_DIRSYNC)
1973 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1975 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0));
1978 static inline int ll_inode_to_ext_flags(int iflags)
1980 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1981 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1982 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1983 #if defined(S_DIRSYNC)
1984 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1986 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0));
1993 struct lustre_handle handle;
1995 __u64 size; /* Offset, in the case of MDS_READPAGE */
1999 __u64 blocks; /* XID, in the case of MDS_READPAGE */
2008 __u32 flags; /* from vfs for pin/unpin, LUSTRE_BFLAG close */
2010 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
2016 __u32 max_cookiesize;
2017 __u32 uid_h; /* high 32-bits of uid, for FUID */
2018 __u32 gid_h; /* high 32-bits of gid, for FUID */
2019 __u32 padding_5; /* also fix lustre_swab_mdt_body */
2027 extern void lustre_swab_mdt_body (struct mdt_body *b);
2029 struct mdt_ioepoch {
2030 struct lustre_handle handle;
2036 extern void lustre_swab_mdt_ioepoch (struct mdt_ioepoch *b);
2038 /* permissions for md_perm.mp_perm */
2040 CFS_SETUID_PERM = 0x01,
2041 CFS_SETGID_PERM = 0x02,
2042 CFS_SETGRP_PERM = 0x04,
2043 CFS_RMTACL_PERM = 0x08,
2044 CFS_RMTOWN_PERM = 0x10
2047 /* inode access permission for remote user, the inode info are omitted,
2048 * for client knows them. */
2049 struct mdt_remote_perm {
2056 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
2060 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
2062 struct mdt_rec_setattr {
2072 __u32 sa_padding_1_h;
2073 struct lu_fid sa_fid;
2082 __u32 sa_attr_flags;
2084 __u32 sa_bias; /* some operation flags */
2090 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
2093 * Attribute flags used in mdt_rec_setattr::sa_valid.
2094 * The kernel's #defines for ATTR_* should not be used over the network
2095 * since the client and MDS may run different kernels (see bug 13828)
2096 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
2098 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
2099 #define MDS_ATTR_UID 0x2ULL /* = 2 */
2100 #define MDS_ATTR_GID 0x4ULL /* = 4 */
2101 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
2102 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
2103 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
2104 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
2105 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
2106 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
2107 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
2108 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
2109 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
2110 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
2111 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
2112 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
2113 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
2116 #define FMODE_READ 00000001
2117 #define FMODE_WRITE 00000002
2120 #define MDS_FMODE_CLOSED 00000000
2121 #define MDS_FMODE_EXEC 00000004
2122 /* IO Epoch is opened on a closed file. */
2123 #define MDS_FMODE_EPOCH 01000000
2124 /* IO Epoch is opened on a file truncate. */
2125 #define MDS_FMODE_TRUNC 02000000
2126 /* Size-on-MDS Attribute Update is pending. */
2127 #define MDS_FMODE_SOM 04000000
2129 #define MDS_OPEN_CREATED 00000010
2130 #define MDS_OPEN_CROSS 00000020
2132 #define MDS_OPEN_CREAT 00000100
2133 #define MDS_OPEN_EXCL 00000200
2134 #define MDS_OPEN_TRUNC 00001000
2135 #define MDS_OPEN_APPEND 00002000
2136 #define MDS_OPEN_SYNC 00010000
2137 #define MDS_OPEN_DIRECTORY 00200000
2139 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
2140 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
2141 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
2142 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
2143 * We do not support JOIN FILE
2144 * anymore, reserve this flags
2145 * just for preventing such bit
2148 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
2149 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
2150 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
2151 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
2152 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
2154 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
2157 /* permission for create non-directory file */
2158 #define MAY_CREATE (1 << 7)
2159 /* permission for create directory file */
2160 #define MAY_LINK (1 << 8)
2161 /* permission for delete from the directory */
2162 #define MAY_UNLINK (1 << 9)
2163 /* source's permission for rename */
2164 #define MAY_RENAME_SRC (1 << 10)
2165 /* target's permission for rename */
2166 #define MAY_RENAME_TAR (1 << 11)
2167 /* part (parent's) VTX permission check */
2168 #define MAY_VTX_PART (1 << 12)
2169 /* full VTX permission check */
2170 #define MAY_VTX_FULL (1 << 13)
2171 /* lfs rgetfacl permission check */
2172 #define MAY_RGETFACL (1 << 14)
2175 MDS_CHECK_SPLIT = 1 << 0,
2176 MDS_CROSS_REF = 1 << 1,
2177 MDS_VTX_BYPASS = 1 << 2,
2178 MDS_PERM_BYPASS = 1 << 3,
2180 MDS_QUOTA_IGNORE = 1 << 5,
2181 MDS_CLOSE_CLEANUP = 1 << 6,
2182 MDS_KEEP_ORPHAN = 1 << 7,
2183 MDS_RECOV_OPEN = 1 << 8,
2184 MDS_DATA_MODIFIED = 1 << 9,
2185 MDS_CREATE_VOLATILE = 1 << 10,
2188 /* instance of mdt_reint_rec */
2189 struct mdt_rec_create {
2197 __u32 cr_suppgid1_h;
2199 __u32 cr_suppgid2_h;
2200 struct lu_fid cr_fid1;
2201 struct lu_fid cr_fid2;
2202 struct lustre_handle cr_old_handle; /* handle in case of open replay */
2206 __u64 cr_padding_1; /* rr_blocks */
2209 /* use of helpers set/get_mrc_cr_flags() is needed to access
2210 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2211 * extend cr_flags size without breaking 1.8 compat */
2212 __u32 cr_flags_l; /* for use with open, low 32 bits */
2213 __u32 cr_flags_h; /* for use with open, high 32 bits */
2214 __u32 cr_umask; /* umask for create */
2215 __u32 cr_padding_4; /* rr_padding_4 */
2218 static inline void set_mrc_cr_flags(struct mdt_rec_create *mrc, __u64 flags)
2220 mrc->cr_flags_l = (__u32)(flags & 0xFFFFFFFFUll);
2221 mrc->cr_flags_h = (__u32)(flags >> 32);
2224 static inline __u64 get_mrc_cr_flags(struct mdt_rec_create *mrc)
2226 return ((__u64)(mrc->cr_flags_l) | ((__u64)mrc->cr_flags_h << 32));
2229 /* instance of mdt_reint_rec */
2230 struct mdt_rec_link {
2238 __u32 lk_suppgid1_h;
2240 __u32 lk_suppgid2_h;
2241 struct lu_fid lk_fid1;
2242 struct lu_fid lk_fid2;
2244 __u64 lk_padding_1; /* rr_atime */
2245 __u64 lk_padding_2; /* rr_ctime */
2246 __u64 lk_padding_3; /* rr_size */
2247 __u64 lk_padding_4; /* rr_blocks */
2249 __u32 lk_padding_5; /* rr_mode */
2250 __u32 lk_padding_6; /* rr_flags */
2251 __u32 lk_padding_7; /* rr_padding_2 */
2252 __u32 lk_padding_8; /* rr_padding_3 */
2253 __u32 lk_padding_9; /* rr_padding_4 */
2256 /* instance of mdt_reint_rec */
2257 struct mdt_rec_unlink {
2265 __u32 ul_suppgid1_h;
2267 __u32 ul_suppgid2_h;
2268 struct lu_fid ul_fid1;
2269 struct lu_fid ul_fid2;
2271 __u64 ul_padding_2; /* rr_atime */
2272 __u64 ul_padding_3; /* rr_ctime */
2273 __u64 ul_padding_4; /* rr_size */
2274 __u64 ul_padding_5; /* rr_blocks */
2277 __u32 ul_padding_6; /* rr_flags */
2278 __u32 ul_padding_7; /* rr_padding_2 */
2279 __u32 ul_padding_8; /* rr_padding_3 */
2280 __u32 ul_padding_9; /* rr_padding_4 */
2283 /* instance of mdt_reint_rec */
2284 struct mdt_rec_rename {
2292 __u32 rn_suppgid1_h;
2294 __u32 rn_suppgid2_h;
2295 struct lu_fid rn_fid1;
2296 struct lu_fid rn_fid2;
2298 __u64 rn_padding_1; /* rr_atime */
2299 __u64 rn_padding_2; /* rr_ctime */
2300 __u64 rn_padding_3; /* rr_size */
2301 __u64 rn_padding_4; /* rr_blocks */
2302 __u32 rn_bias; /* some operation flags */
2303 __u32 rn_mode; /* cross-ref rename has mode */
2304 __u32 rn_padding_5; /* rr_flags */
2305 __u32 rn_padding_6; /* rr_padding_2 */
2306 __u32 rn_padding_7; /* rr_padding_3 */
2307 __u32 rn_padding_8; /* rr_padding_4 */
2310 /* instance of mdt_reint_rec */
2311 struct mdt_rec_setxattr {
2319 __u32 sx_suppgid1_h;
2321 __u32 sx_suppgid2_h;
2322 struct lu_fid sx_fid;
2323 __u64 sx_padding_1; /* These three are rr_fid2 */
2328 __u64 sx_padding_5; /* rr_ctime */
2329 __u64 sx_padding_6; /* rr_size */
2330 __u64 sx_padding_7; /* rr_blocks */
2333 __u32 sx_padding_8; /* rr_flags */
2334 __u32 sx_padding_9; /* rr_padding_2 */
2335 __u32 sx_padding_10; /* rr_padding_3 */
2336 __u32 sx_padding_11; /* rr_padding_4 */
2340 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2341 * Do NOT change the size of various members, otherwise the value
2342 * will be broken in lustre_swab_mdt_rec_reint().
2344 * If you add new members in other mdt_reint_xxx structres and need to use the
2345 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2347 struct mdt_rec_reint {
2355 __u32 rr_suppgid1_h;
2357 __u32 rr_suppgid2_h;
2358 struct lu_fid rr_fid1;
2359 struct lu_fid rr_fid2;
2368 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
2369 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
2370 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2373 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
2376 __u32 ld_tgt_count; /* how many MDS's */
2377 __u32 ld_active_tgt_count; /* how many active */
2378 __u32 ld_default_stripe_count; /* how many objects are used */
2379 __u32 ld_pattern; /* default MEA_MAGIC_* */
2380 __u64 ld_default_hash_size;
2381 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2382 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2383 __u32 ld_qos_maxage; /* in second */
2384 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2385 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2386 struct obd_uuid ld_uuid;
2389 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
2391 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
2392 struct lmv_stripe_md {
2397 char mea_pool_name[LOV_MAXPOOLNAME];
2398 struct lu_fid mea_ids[0];
2401 extern void lustre_swab_lmv_stripe_md(struct lmv_stripe_md *mea);
2403 /* lmv structures */
2404 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
2405 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
2406 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
2408 #define MAX_HASH_SIZE_32 0x7fffffffUL
2409 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
2410 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
2415 FLD_FIRST_OPC = FLD_QUERY
2421 SEQ_FIRST_OPC = SEQ_QUERY
2425 SEQ_ALLOC_SUPER = 0,
2430 * LOV data structures
2433 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2434 /* The size of the buffer the lov/mdc reserves for the
2435 * array of UUIDs returned by the MDS. With the current
2436 * protocol, this will limit the max number of OSTs per LOV */
2438 #define LOV_DESC_MAGIC 0xB0CCDE5C
2440 /* LOV settings descriptor (should only contain static info) */
2442 __u32 ld_tgt_count; /* how many OBD's */
2443 __u32 ld_active_tgt_count; /* how many active */
2444 __u32 ld_default_stripe_count; /* how many objects are used */
2445 __u32 ld_pattern; /* default PATTERN_RAID0 */
2446 __u64 ld_default_stripe_size; /* in bytes */
2447 __u64 ld_default_stripe_offset; /* in bytes */
2448 __u32 ld_padding_0; /* unused */
2449 __u32 ld_qos_maxage; /* in second */
2450 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2451 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2452 struct obd_uuid ld_uuid;
2455 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2457 extern void lustre_swab_lov_desc (struct lov_desc *ld);
2462 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2467 LDLM_BL_CALLBACK = 104,
2468 LDLM_CP_CALLBACK = 105,
2469 LDLM_GL_CALLBACK = 106,
2470 LDLM_SET_INFO = 107,
2473 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2475 #define RES_NAME_SIZE 4
2476 struct ldlm_res_id {
2477 __u64 name[RES_NAME_SIZE];
2480 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
2482 static inline int ldlm_res_eq(const struct ldlm_res_id *res0,
2483 const struct ldlm_res_id *res1)
2485 return !memcmp(res0, res1, sizeof(*res0));
2502 #define LCK_MODE_NUM 8
2512 #define LDLM_MIN_TYPE LDLM_PLAIN
2514 struct ldlm_extent {
2520 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
2521 struct ldlm_extent *ex2)
2523 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
2526 /* check if @ex1 contains @ex2 */
2527 static inline int ldlm_extent_contain(struct ldlm_extent *ex1,
2528 struct ldlm_extent *ex2)
2530 return (ex1->start <= ex2->start) && (ex1->end >= ex2->end);
2533 struct ldlm_inodebits {
2537 struct ldlm_flock_wire {
2545 /* it's important that the fields of the ldlm_extent structure match
2546 * the first fields of the ldlm_flock structure because there is only
2547 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2548 * this ever changes we will need to swab the union differently based
2549 * on the resource type. */
2552 struct ldlm_extent l_extent;
2553 struct ldlm_flock_wire l_flock;
2554 struct ldlm_inodebits l_inodebits;
2555 } ldlm_wire_policy_data_t;
2557 extern void lustre_swab_ldlm_policy_data (ldlm_wire_policy_data_t *d);
2559 union ldlm_gl_desc {
2560 struct ldlm_gl_lquota_desc lquota_desc;
2563 extern void lustre_swab_gl_desc(union ldlm_gl_desc *);
2565 struct ldlm_intent {
2569 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
2571 struct ldlm_resource_desc {
2572 ldlm_type_t lr_type;
2573 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
2574 struct ldlm_res_id lr_name;
2577 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
2579 struct ldlm_lock_desc {
2580 struct ldlm_resource_desc l_resource;
2581 ldlm_mode_t l_req_mode;
2582 ldlm_mode_t l_granted_mode;
2583 ldlm_wire_policy_data_t l_policy_data;
2586 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
2588 #define LDLM_LOCKREQ_HANDLES 2
2589 #define LDLM_ENQUEUE_CANCEL_OFF 1
2591 struct ldlm_request {
2594 struct ldlm_lock_desc lock_desc;
2595 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2598 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
2600 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
2601 * Otherwise, 2 are available. */
2602 #define ldlm_request_bufsize(count,type) \
2604 int _avail = LDLM_LOCKREQ_HANDLES; \
2605 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
2606 sizeof(struct ldlm_request) + \
2607 (count > _avail ? count - _avail : 0) * \
2608 sizeof(struct lustre_handle); \
2613 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2614 struct ldlm_lock_desc lock_desc;
2615 struct lustre_handle lock_handle;
2616 __u64 lock_policy_res1;
2617 __u64 lock_policy_res2;
2620 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
2622 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2623 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2626 * Opcodes for mountconf (mgs and mgc)
2631 MGS_EXCEPTION, /* node died, etc. */
2632 MGS_TARGET_REG, /* whenever target starts up */
2638 #define MGS_FIRST_OPC MGS_CONNECT
2640 #define MGS_PARAM_MAXLEN 1024
2641 #define KEY_SET_INFO "set_info"
2643 struct mgs_send_param {
2644 char mgs_param[MGS_PARAM_MAXLEN];
2647 /* We pass this info to the MGS so it can write config logs */
2648 #define MTI_NAME_MAXLEN 64
2649 #define MTI_PARAM_MAXLEN 4096
2650 #define MTI_NIDS_MAX 32
2651 struct mgs_target_info {
2652 __u32 mti_lustre_ver;
2653 __u32 mti_stripe_index;
2654 __u32 mti_config_ver;
2656 __u32 mti_nid_count;
2657 __u32 mti_instance; /* Running instance of target */
2658 char mti_fsname[MTI_NAME_MAXLEN];
2659 char mti_svname[MTI_NAME_MAXLEN];
2660 char mti_uuid[sizeof(struct obd_uuid)];
2661 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
2662 char mti_params[MTI_PARAM_MAXLEN];
2664 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
2666 struct mgs_nidtbl_entry {
2667 __u64 mne_version; /* table version of this entry */
2668 __u32 mne_instance; /* target instance # */
2669 __u32 mne_index; /* target index */
2670 __u32 mne_length; /* length of this entry - by bytes */
2671 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2672 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2673 __u8 mne_nid_size; /* size of each NID, by bytes */
2674 __u8 mne_nid_count; /* # of NIDs in buffer */
2676 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2679 extern void lustre_swab_mgs_nidtbl_entry(struct mgs_nidtbl_entry *oinfo);
2681 struct mgs_config_body {
2682 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2683 __u64 mcb_offset; /* next index of config log to request */
2684 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2686 __u8 mcb_bits; /* bits unit size of config log */
2687 __u32 mcb_units; /* # of units for bulk transfer */
2689 extern void lustre_swab_mgs_config_body(struct mgs_config_body *body);
2691 struct mgs_config_res {
2692 __u64 mcr_offset; /* index of last config log */
2693 __u64 mcr_size; /* size of the log */
2695 extern void lustre_swab_mgs_config_res(struct mgs_config_res *body);
2697 /* Config marker flags (in config log) */
2698 #define CM_START 0x01
2700 #define CM_SKIP 0x04
2701 #define CM_UPGRADE146 0x08
2702 #define CM_EXCLUDE 0x10
2703 #define CM_START_SKIP (CM_START | CM_SKIP)
2706 __u32 cm_step; /* aka config version */
2708 __u32 cm_vers; /* lustre release version number */
2709 __u32 cm_padding; /* 64 bit align */
2710 obd_time cm_createtime; /*when this record was first created */
2711 obd_time cm_canceltime; /*when this record is no longer valid*/
2712 char cm_tgtname[MTI_NAME_MAXLEN];
2713 char cm_comment[MTI_NAME_MAXLEN];
2716 extern void lustre_swab_cfg_marker(struct cfg_marker *marker,
2717 int swab, int size);
2720 * Opcodes for multiple servers.
2730 #define OBD_FIRST_OPC OBD_PING
2732 /* catalog of log objects */
2734 /** Identifier for a single log object */
2739 } __attribute__((packed));
2741 /** Records written to the CATALOGS list */
2742 #define CATLIST "CATALOGS"
2744 struct llog_logid lci_logid;
2748 } __attribute__((packed));
2750 /* Log data record types - there is no specific reason that these need to
2751 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2753 #define LLOG_OP_MAGIC 0x10600000
2754 #define LLOG_OP_MASK 0xfff00000
2757 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2758 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2759 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2760 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2761 REINT_UNLINK, /* obsolete after 2.5.0 */
2762 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2764 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2765 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2767 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2768 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2769 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2770 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2771 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2772 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2773 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2774 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2777 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2778 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2780 /** Log record header - stored in little endian order.
2781 * Each record must start with this struct, end with a llog_rec_tail,
2782 * and be a multiple of 256 bits in size.
2784 struct llog_rec_hdr {
2791 struct llog_rec_tail {
2796 /* Where data follow just after header */
2797 #define REC_DATA(ptr) \
2798 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2800 #define REC_DATA_LEN(rec) \
2801 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2802 sizeof(struct llog_rec_tail))
2804 struct llog_logid_rec {
2805 struct llog_rec_hdr lid_hdr;
2806 struct llog_logid lid_id;
2810 struct llog_rec_tail lid_tail;
2811 } __attribute__((packed));
2813 struct llog_unlink_rec {
2814 struct llog_rec_hdr lur_hdr;
2817 obd_count lur_count;
2818 struct llog_rec_tail lur_tail;
2819 } __attribute__((packed));
2821 struct llog_unlink64_rec {
2822 struct llog_rec_hdr lur_hdr;
2823 struct lu_fid lur_fid;
2824 obd_count lur_count; /* to destroy the lost precreated */
2828 struct llog_rec_tail lur_tail;
2829 } __attribute__((packed));
2831 struct llog_setattr64_rec {
2832 struct llog_rec_hdr lsr_hdr;
2840 struct llog_rec_tail lsr_tail;
2841 } __attribute__((packed));
2843 struct llog_size_change_rec {
2844 struct llog_rec_hdr lsc_hdr;
2845 struct ll_fid lsc_fid;
2850 struct llog_rec_tail lsc_tail;
2851 } __attribute__((packed));
2853 #define CHANGELOG_MAGIC 0xca103000
2855 /** \a changelog_rec_type's that can't be masked */
2856 #define CHANGELOG_MINMASK (1 << CL_MARK)
2857 /** bits covering all \a changelog_rec_type's */
2858 #define CHANGELOG_ALLMASK 0XFFFFFFFF
2859 /** default \a changelog_rec_type mask */
2860 #define CHANGELOG_DEFMASK CHANGELOG_ALLMASK & ~(1 << CL_ATIME | 1 << CL_CLOSE)
2862 /* changelog llog name, needed by client replicators */
2863 #define CHANGELOG_CATALOG "changelog_catalog"
2865 struct changelog_setinfo {
2868 } __attribute__((packed));
2870 /** changelog record */
2871 struct llog_changelog_rec {
2872 struct llog_rec_hdr cr_hdr;
2873 struct changelog_rec cr;
2874 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
2875 } __attribute__((packed));
2877 struct llog_changelog_ext_rec {
2878 struct llog_rec_hdr cr_hdr;
2879 struct changelog_ext_rec cr;
2880 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
2881 } __attribute__((packed));
2883 #define CHANGELOG_USER_PREFIX "cl"
2885 struct llog_changelog_user_rec {
2886 struct llog_rec_hdr cur_hdr;
2890 struct llog_rec_tail cur_tail;
2891 } __attribute__((packed));
2893 /* Old llog gen for compatibility */
2897 } __attribute__((packed));
2899 struct llog_gen_rec {
2900 struct llog_rec_hdr lgr_hdr;
2901 struct llog_gen lgr_gen;
2905 struct llog_rec_tail lgr_tail;
2908 /* On-disk header structure of each log object, stored in little endian order */
2909 #define LLOG_CHUNK_SIZE 8192
2910 #define LLOG_HEADER_SIZE (96)
2911 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2913 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2915 /* flags for the logs */
2917 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2918 LLOG_F_IS_CAT = 0x2,
2919 LLOG_F_IS_PLAIN = 0x4,
2922 struct llog_log_hdr {
2923 struct llog_rec_hdr llh_hdr;
2924 obd_time llh_timestamp;
2926 __u32 llh_bitmap_offset;
2930 /* for a catalog the first plain slot is next to it */
2931 struct obd_uuid llh_tgtuuid;
2932 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2933 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2934 struct llog_rec_tail llh_tail;
2935 } __attribute__((packed));
2937 #define LLOG_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2938 llh->llh_bitmap_offset - \
2939 sizeof(llh->llh_tail)) * 8)
2941 /** log cookies are used to reference a specific log file and a record therein */
2942 struct llog_cookie {
2943 struct llog_logid lgc_lgl;
2947 } __attribute__((packed));
2949 /** llog protocol */
2950 enum llogd_rpc_ops {
2951 LLOG_ORIGIN_HANDLE_CREATE = 501,
2952 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2953 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2954 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2955 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2956 LLOG_ORIGIN_CONNECT = 506,
2957 LLOG_CATINFO = 507, /* deprecated */
2958 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2959 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2961 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2965 struct llog_logid lgd_logid;
2967 __u32 lgd_llh_flags;
2969 __u32 lgd_saved_index;
2971 __u64 lgd_cur_offset;
2972 } __attribute__((packed));
2974 struct llogd_conn_body {
2975 struct llog_gen lgdc_gen;
2976 struct llog_logid lgdc_logid;
2977 __u32 lgdc_ctxt_idx;
2978 } __attribute__((packed));
2980 /* Note: 64-bit types are 64-bit aligned in structure */
2982 obd_valid o_valid; /* hot fields in this obdo */
2984 obd_id o_parent_seq;
2985 obd_size o_size; /* o_size-o_blocks == ost_lvb */
2989 obd_blocks o_blocks; /* brw: cli sent cached bytes */
2992 /* 32-bit fields start here: keep an even number of them via padding */
2993 obd_blksize o_blksize; /* optimal IO blocksize */
2994 obd_mode o_mode; /* brw: cli sent cache remain */
2998 obd_count o_nlink; /* brw: checksum */
2999 obd_count o_parent_oid;
3000 obd_count o_misc; /* brw: o_dropped */
3001 __u64 o_ioepoch; /* epoch in ost writes */
3002 __u32 o_stripe_idx; /* holds stripe idx */
3004 struct lustre_handle o_handle; /* brw: lock handle to prolong
3006 struct llog_cookie o_lcookie; /* destroy: unlink cookie from
3011 __u64 o_data_version; /* getattr: sum of iversion for
3013 * brw: grant space consumed on
3014 * the client for the write */
3020 #define o_id o_oi.oi_id
3021 #define o_seq o_oi.oi_seq
3022 #define o_dirty o_blocks
3023 #define o_undirty o_mode
3024 #define o_dropped o_misc
3025 #define o_cksum o_nlink
3026 #define o_grant_used o_data_version
3028 static inline void lustre_set_wire_obdo(struct obdo *wobdo, struct obdo *lobdo)
3030 memcpy(wobdo, lobdo, sizeof(*lobdo));
3031 wobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3034 static inline void lustre_get_wire_obdo(struct obdo *lobdo, struct obdo *wobdo)
3036 obd_flag local_flags = 0;
3038 if (lobdo->o_valid & OBD_MD_FLFLAGS)
3039 local_flags = lobdo->o_flags & OBD_FL_LOCAL_MASK;
3041 LASSERT(!(wobdo->o_flags & OBD_FL_LOCAL_MASK));
3043 memcpy(lobdo, wobdo, sizeof(*lobdo));
3044 if (local_flags != 0) {
3045 lobdo->o_valid |= OBD_MD_FLFLAGS;
3046 lobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3047 lobdo->o_flags |= local_flags;
3051 extern void lustre_swab_obdo (struct obdo *o);
3053 /* request structure for OST's */
3058 /* Key for FIEMAP to be used in get_info calls */
3059 struct ll_fiemap_info_key {
3062 struct ll_user_fiemap fiemap;
3065 extern void lustre_swab_ost_body (struct ost_body *b);
3066 extern void lustre_swab_ost_last_id(obd_id *id);
3067 extern void lustre_swab_fiemap(struct ll_user_fiemap *fiemap);
3069 extern void lustre_swab_lov_user_md_v1(struct lov_user_md_v1 *lum);
3070 extern void lustre_swab_lov_user_md_v3(struct lov_user_md_v3 *lum);
3071 extern void lustre_swab_lov_user_md_objects(struct lov_user_ost_data *lod,
3073 extern void lustre_swab_lov_mds_md(struct lov_mds_md *lmm);
3076 extern void lustre_swab_llogd_body (struct llogd_body *d);
3077 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
3078 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
3079 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec);
3082 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
3084 /* Functions for dumping PTLRPC fields */
3085 void dump_rniobuf(struct niobuf_remote *rnb);
3086 void dump_ioo(struct obd_ioobj *nb);
3087 void dump_obdo(struct obdo *oa);
3088 void dump_ost_body(struct ost_body *ob);
3089 void dump_rcs(__u32 *rc);
3091 #define IDX_INFO_MAGIC 0x3D37CC37
3093 /* Index file transfer through the network. The server serializes the index into
3094 * a byte stream which is sent to the client via a bulk transfer */
3098 /* reply: see idx_info_flags below */
3101 /* request & reply: number of lu_idxpage (to be) transferred */
3105 /* request: requested attributes passed down to the iterator API */
3108 /* request & reply: index file identifier (FID) */
3109 struct lu_fid ii_fid;
3111 /* reply: version of the index file before starting to walk the index.
3112 * Please note that the version can be modified at any time during the
3116 /* request: hash to start with:
3117 * reply: hash of the first entry of the first lu_idxpage and hash
3118 * of the entry to read next if any */
3119 __u64 ii_hash_start;
3122 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3126 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3134 extern void lustre_swab_idx_info(struct idx_info *ii);
3136 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3138 /* List of flags used in idx_info::ii_flags */
3139 enum idx_info_flags {
3140 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3141 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3142 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3143 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3146 #define LIP_MAGIC 0x8A6D6B6C
3148 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3150 /* 16-byte header */
3153 __u16 lip_nr; /* number of entries in the container */
3154 __u64 lip_pad0; /* additional padding for future use */
3156 /* key/record pairs are stored in the remaining 4080 bytes.
3157 * depending upon the flags in idx_info::ii_flags, each key/record
3158 * pair might be preceded by:
3160 * - the key size (II_FL_VARKEY is set)
3161 * - the record size (II_FL_VARREC is set)
3163 * For the time being, we only support fixed-size key & record. */
3164 char lip_entries[0];
3166 extern void lustre_swab_lip_header(struct lu_idxpage *lip);
3168 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3170 /* Gather all possible type associated with a 4KB container */
3172 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3173 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3174 char lp_array[LU_PAGE_SIZE];
3177 /* security opcodes */
3180 SEC_CTX_INIT_CONT = 802,
3183 SEC_FIRST_OPC = SEC_CTX_INIT
3187 * capa related definitions
3189 #define CAPA_HMAC_MAX_LEN 64
3190 #define CAPA_HMAC_KEY_MAX_LEN 56
3192 /* NB take care when changing the sequence of elements this struct,
3193 * because the offset info is used in find_capa() */
3194 struct lustre_capa {
3195 struct lu_fid lc_fid; /** fid */
3196 __u64 lc_opc; /** operations allowed */
3197 __u64 lc_uid; /** file owner */
3198 __u64 lc_gid; /** file group */
3199 __u32 lc_flags; /** HMAC algorithm & flags */
3200 __u32 lc_keyid; /** key# used for the capability */
3201 __u32 lc_timeout; /** capa timeout value (sec) */
3202 __u32 lc_expiry; /** expiry time (sec) */
3203 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
3204 } __attribute__((packed));
3206 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
3208 /** lustre_capa::lc_opc */
3210 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
3211 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
3212 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
3213 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
3214 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
3215 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
3216 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
3217 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
3218 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
3219 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
3220 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
3223 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3224 #define CAPA_OPC_MDS_ONLY \
3225 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3226 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3227 #define CAPA_OPC_OSS_ONLY \
3228 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3229 CAPA_OPC_OSS_DESTROY)
3230 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3231 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3233 /* MDS capability covers object capability for operations of body r/w
3234 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
3235 * while OSS capability only covers object capability for operations of
3236 * oss data(file content) r/w/truncate.
3238 static inline int capa_for_mds(struct lustre_capa *c)
3240 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
3243 static inline int capa_for_oss(struct lustre_capa *c)
3245 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
3248 /* lustre_capa::lc_hmac_alg */
3250 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
3254 #define CAPA_FL_MASK 0x00ffffff
3255 #define CAPA_HMAC_ALG_MASK 0xff000000
3257 struct lustre_capa_key {
3258 __u64 lk_seq; /**< mds# */
3259 __u32 lk_keyid; /**< key# */
3261 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
3262 } __attribute__((packed));
3264 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
3266 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3267 #define LINK_EA_MAGIC 0x11EAF1DFUL
3268 struct link_ea_header {
3271 __u64 leh_len; /* total size */
3277 /** Hardlink data is name and parent fid.
3278 * Stored in this crazy struct for maximum packing and endian-neutrality
3280 struct link_ea_entry {
3281 /** __u16 stored big-endian, unaligned */
3282 unsigned char lee_reclen[2];
3283 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3285 }__attribute__((packed));
3287 /** fid2path request/reply structure */
3288 struct getinfo_fid2path {
3289 struct lu_fid gf_fid;
3294 } __attribute__((packed));
3296 void lustre_swab_fid2path (struct getinfo_fid2path *gf);
3299 LAYOUT_INTENT_ACCESS = 0,
3300 LAYOUT_INTENT_READ = 1,
3301 LAYOUT_INTENT_WRITE = 2,
3302 LAYOUT_INTENT_GLIMPSE = 3,
3303 LAYOUT_INTENT_TRUNC = 4,
3304 LAYOUT_INTENT_RELEASE = 5,
3305 LAYOUT_INTENT_RESTORE = 6
3308 /* enqueue layout lock with intent */
3309 struct layout_intent {
3310 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3316 void lustre_swab_layout_intent(struct layout_intent *li);
3319 * On the wire version of hsm_progress structure.
3321 * Contains the userspace hsm_progress and some internal fields.
3323 struct hsm_progress_kernel {
3324 /* Field taken from struct hsm_progress */
3327 struct hsm_extent hpk_extent;
3329 __u16 hpk_errval; /* positive val */
3331 /* Additional fields */
3332 __u64 hpk_data_version;
3334 } __attribute__((packed));
3336 extern void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
3337 extern void lustre_swab_hsm_current_action(struct hsm_current_action *action);
3338 extern void lustre_swab_hsm_progress_kernel(struct hsm_progress_kernel *hpk);
3339 extern void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
3340 extern void lustre_swab_hsm_user_item(struct hsm_user_item *hui);
3341 extern void lustre_swab_hsm_request(struct hsm_request *hr);