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);
357 /* This is the maximum number of MDTs allowed in CMD testing until such
358 * a time that FID-on-OST is implemented. This is due to the limitations
359 * of packing non-0-MDT numbers into the FID SEQ namespace. Once FID-on-OST
360 * is implemented this limit will be virtually unlimited. */
361 #define MAX_MDT_COUNT 8
368 /** initial fid id value */
369 LUSTRE_FID_INIT_OID = 1UL
372 /** returns fid object sequence */
373 static inline __u64 fid_seq(const struct lu_fid *fid)
378 /** returns fid object id */
379 static inline __u32 fid_oid(const struct lu_fid *fid)
384 /** returns fid object version */
385 static inline __u32 fid_ver(const struct lu_fid *fid)
390 static inline void fid_zero(struct lu_fid *fid)
392 memset(fid, 0, sizeof(*fid));
395 static inline obd_id fid_ver_oid(const struct lu_fid *fid)
397 return ((__u64)fid_ver(fid) << 32 | fid_oid(fid));
401 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
402 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
403 * used for other purposes and not risk collisions with existing inodes.
405 * Different FID Format
406 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
409 FID_SEQ_OST_MDT0 = 0,
412 FID_SEQ_OST_MDT1 = 3,
413 FID_SEQ_OST_MAX = 9, /* Max MDT count before OST_on_FID */
416 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
417 FID_SEQ_IDIF = 0x100000000ULL,
418 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
419 /* Normal FID sequence starts from this value, i.e. 1<<33 */
420 FID_SEQ_START = 0x200000000ULL,
421 /* sequence for local pre-defined FIDs listed in local_oid */
422 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
423 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
424 /* sequence is used for local named objects FIDs generated
425 * by local_object_storage library */
426 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
427 FID_SEQ_SPECIAL = 0x200000004ULL,
428 FID_SEQ_QUOTA = 0x200000005ULL,
429 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
430 FID_SEQ_NORMAL = 0x200000400ULL,
431 FID_SEQ_LOV_DEFAULT= 0xffffffffffffffffULL
434 #define OBIF_OID_MAX_BITS 32
435 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
436 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
437 #define IDIF_OID_MAX_BITS 48
438 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
439 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
441 /** OID for FID_SEQ_SPECIAL */
443 /* Big Filesystem Lock to serialize rename operations */
444 FID_OID_SPECIAL_BFL = 1UL,
447 /** OID for FID_SEQ_DOT_LUSTRE */
448 enum dot_lustre_oid {
449 FID_OID_DOT_LUSTRE = 1UL,
450 FID_OID_DOT_LUSTRE_OBF = 2UL,
453 static inline int fid_seq_is_mdt0(obd_seq seq)
455 return (seq == FID_SEQ_OST_MDT0);
458 static inline int fid_seq_is_mdt(const __u64 seq)
460 return seq == FID_SEQ_OST_MDT0 || seq >= FID_SEQ_NORMAL;
463 static inline int fid_seq_is_echo(obd_seq seq)
465 return (seq == FID_SEQ_ECHO);
468 static inline int fid_is_echo(const struct lu_fid *fid)
470 return fid_seq_is_echo(fid_seq(fid));
473 static inline int fid_seq_is_llog(obd_seq seq)
475 return (seq == FID_SEQ_LLOG);
478 static inline int fid_is_llog(const struct lu_fid *fid)
480 return fid_seq_is_llog(fid_seq(fid));
483 static inline int fid_seq_is_rsvd(const __u64 seq)
485 return (seq > FID_SEQ_OST_MDT0 && seq <= FID_SEQ_RSVD);
488 static inline int fid_is_mdt0(const struct lu_fid *fid)
490 return fid_seq_is_mdt0(fid_seq(fid));
494 * Check if a fid is igif or not.
495 * \param fid the fid to be tested.
496 * \return true if the fid is a igif; otherwise false.
498 static inline int fid_seq_is_igif(const __u64 seq)
500 return seq >= FID_SEQ_IGIF && seq <= FID_SEQ_IGIF_MAX;
503 static inline int fid_is_igif(const struct lu_fid *fid)
505 return fid_seq_is_igif(fid_seq(fid));
509 * Check if a fid is idif or not.
510 * \param fid the fid to be tested.
511 * \return true if the fid is a idif; otherwise false.
513 static inline int fid_seq_is_idif(const __u64 seq)
515 return seq >= FID_SEQ_IDIF && seq <= FID_SEQ_IDIF_MAX;
518 static inline int fid_is_idif(const struct lu_fid *fid)
520 return fid_seq_is_idif(fid_seq(fid));
528 static inline int fid_seq_is_norm(const __u64 seq)
530 return (seq >= FID_SEQ_NORMAL);
533 static inline int fid_is_norm(const struct lu_fid *fid)
535 return fid_seq_is_norm(fid_seq(fid));
538 /* convert an OST objid into an IDIF FID SEQ number */
539 static inline obd_seq fid_idif_seq(obd_id id, __u32 ost_idx)
541 return FID_SEQ_IDIF | (ost_idx << 16) | ((id >> 32) & 0xffff);
544 /* convert a packed IDIF FID into an OST objid */
545 static inline obd_id fid_idif_id(obd_seq seq, __u32 oid, __u32 ver)
547 return ((__u64)ver << 48) | ((seq & 0xffff) << 32) | oid;
550 /* extract ost index from IDIF FID */
551 static inline __u32 fid_idif_ost_idx(const struct lu_fid *fid)
553 LASSERT(fid_is_idif(fid));
554 return (fid_seq(fid) >> 16) & 0xffff;
557 /* unpack an ostid (id/seq) from a wire/disk structure into an IDIF FID */
558 static inline void ostid_idif_unpack(struct ost_id *ostid,
559 struct lu_fid *fid, __u32 ost_idx)
561 fid->f_seq = fid_idif_seq(ostid->oi_id, ost_idx);
562 fid->f_oid = ostid->oi_id; /* truncate to 32 bits by assignment */
563 fid->f_ver = ostid->oi_id >> 48; /* in theory, not currently used */
566 /* unpack an ostid (id/seq) from a wire/disk structure into a non-IDIF FID */
567 static inline void ostid_fid_unpack(struct ost_id *ostid, struct lu_fid *fid)
569 fid->f_seq = ostid->oi_seq;
570 fid->f_oid = ostid->oi_id; /* truncate to 32 bits by assignment */
571 fid->f_ver = ostid->oi_id >> 32; /* in theory, not currently used */
574 /* Unpack an OST object id/seq (group) into a FID. This is needed for
575 * converting all obdo, lmm, lsm, etc. 64-bit id/seq pairs into proper
576 * FIDs. Note that if an id/seq is already in FID/IDIF format it will
577 * be passed through unchanged. Only legacy OST objects in "group 0"
578 * will be mapped into the IDIF namespace so that they can fit into the
579 * struct lu_fid fields without loss. For reference see:
580 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs
582 static inline int fid_ostid_unpack(struct lu_fid *fid, struct ost_id *ostid,
585 if (ost_idx > 0xffff) {
586 CERROR("bad ost_idx, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
587 ostid->oi_seq, ostid->oi_id, ost_idx);
591 if (fid_seq_is_mdt0(ostid->oi_seq)) {
592 /* This is a "legacy" (old 1.x/2.early) OST object in "group 0"
593 * that we map into the IDIF namespace. It allows up to 2^48
594 * objects per OST, as this is the object namespace that has
595 * been in production for years. This can handle create rates
596 * of 1M objects/s/OST for 9 years, or combinations thereof. */
597 if (ostid->oi_id >= IDIF_MAX_OID) {
598 CERROR("bad MDT0 id, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
599 ostid->oi_seq, ostid->oi_id, ost_idx);
602 ostid_idif_unpack(ostid, fid, ost_idx);
604 } else if (fid_seq_is_rsvd(ostid->oi_seq)) {
605 /* These are legacy OST objects for LLOG/ECHO and CMD testing.
606 * We only support 2^32 objects in these groups, and cannot
607 * uniquely identify them in the system (i.e. they are the
608 * duplicated on all OSTs), but this is not strictly required
609 * for the old object protocol, which has a separate ost_idx. */
610 if (ostid->oi_id >= 0xffffffffULL) {
611 CERROR("bad RSVD id, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
612 ostid->oi_seq, ostid->oi_id, ost_idx);
615 ostid_fid_unpack(ostid, fid);
617 } else if (unlikely(fid_seq_is_igif(ostid->oi_seq))) {
618 /* This is an MDT inode number, which should never collide with
619 * proper OST object IDs, and is probably a broken filesystem */
620 CERROR("bad IGIF, seq:"LPU64" id:"LPU64" ost_idx:%u\n",
621 ostid->oi_seq, ostid->oi_id, ost_idx);
624 } else /* if (fid_seq_is_idif(seq) || fid_seq_is_norm(seq)) */ {
625 /* This is either an IDIF object, which identifies objects across
626 * all OSTs, or a regular FID. The IDIF namespace maps legacy
627 * OST objects into the FID namespace. In both cases, we just
628 * pass the FID through, no conversion needed. */
629 ostid_fid_unpack(ostid, fid);
635 /* pack an IDIF FID into an ostid (id/seq) for the wire/disk */
636 static inline void ostid_idif_pack(const struct lu_fid *fid,
637 struct ost_id *ostid)
639 ostid->oi_seq = FID_SEQ_OST_MDT0;
640 ostid->oi_id = fid_idif_id(fid->f_seq, fid->f_oid, fid->f_ver);
643 /* pack a non-IDIF FID into an ostid (id/seq) for the wire/disk */
644 static inline void ostid_fid_pack(const struct lu_fid *fid,
645 struct ost_id *ostid)
647 ostid->oi_seq = fid_seq(fid);
648 ostid->oi_id = fid_ver_oid(fid);
651 /* pack any OST FID into an ostid (id/seq) for the wire/disk */
652 static inline int fid_ostid_pack(const struct lu_fid *fid,
653 struct ost_id *ostid)
655 if (unlikely(fid_seq_is_igif(fid->f_seq))) {
656 CERROR("bad IGIF, "DFID"\n", PFID(fid));
660 if (fid_is_idif(fid))
661 ostid_idif_pack(fid, ostid);
663 ostid_fid_pack(fid, ostid);
668 /* extract OST sequence (group) from a wire ost_id (id/seq) pair */
669 static inline obd_seq ostid_seq(struct ost_id *ostid)
671 if (unlikely(fid_seq_is_igif(ostid->oi_seq)))
672 CWARN("bad IGIF, oi_seq: "LPU64" oi_id: "LPX64"\n",
673 ostid->oi_seq, ostid->oi_id);
675 if (unlikely(fid_seq_is_idif(ostid->oi_seq)))
676 return FID_SEQ_OST_MDT0;
678 return ostid->oi_seq;
681 /* extract OST objid from a wire ost_id (id/seq) pair */
682 static inline obd_id ostid_id(struct ost_id *ostid)
684 if (ostid->oi_seq == FID_SEQ_OST_MDT0)
685 return ostid->oi_id & IDIF_OID_MASK;
687 if (fid_seq_is_rsvd(ostid->oi_seq))
688 return ostid->oi_id & OBIF_OID_MASK;
690 if (fid_seq_is_idif(ostid->oi_seq))
691 return fid_idif_id(ostid->oi_seq, ostid->oi_id, 0);
696 /* Check whether the fid is for LAST_ID */
697 static inline int fid_is_last_id(const struct lu_fid *fid)
699 return (fid_is_idif(fid) || fid_is_norm(fid) || fid_is_echo(fid)) &&
704 * Get inode number from a igif.
705 * \param fid a igif to get inode number from.
706 * \return inode number for the igif.
708 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
714 * Build igif from the inode number/generation.
716 #define LU_IGIF_BUILD(fid, ino, gen) \
722 static inline void lu_igif_build(struct lu_fid *fid, __u32 ino, __u32 gen)
724 LU_IGIF_BUILD(fid, ino, gen);
725 LASSERT(fid_is_igif(fid));
729 * Get inode generation from a igif.
730 * \param fid a igif to get inode generation from.
731 * \return inode generation for the igif.
733 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
739 * Fids are transmitted across network (in the sender byte-ordering),
740 * and stored on disk in big-endian order.
742 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
744 /* check that all fields are converted */
745 CLASSERT(sizeof *src ==
746 sizeof fid_seq(src) +
747 sizeof fid_oid(src) + sizeof fid_ver(src));
748 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
749 dst->f_seq = cpu_to_le64(fid_seq(src));
750 dst->f_oid = cpu_to_le32(fid_oid(src));
751 dst->f_ver = cpu_to_le32(fid_ver(src));
754 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
756 /* check that all fields are converted */
757 CLASSERT(sizeof *src ==
758 sizeof fid_seq(src) +
759 sizeof fid_oid(src) + sizeof fid_ver(src));
760 dst->f_seq = le64_to_cpu(fid_seq(src));
761 dst->f_oid = le32_to_cpu(fid_oid(src));
762 dst->f_ver = le32_to_cpu(fid_ver(src));
763 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
766 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
768 /* check that all fields are converted */
769 CLASSERT(sizeof *src ==
770 sizeof fid_seq(src) +
771 sizeof fid_oid(src) + sizeof fid_ver(src));
772 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
773 dst->f_seq = cpu_to_be64(fid_seq(src));
774 dst->f_oid = cpu_to_be32(fid_oid(src));
775 dst->f_ver = cpu_to_be32(fid_ver(src));
778 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
780 /* check that all fields are converted */
781 CLASSERT(sizeof *src ==
782 sizeof fid_seq(src) +
783 sizeof fid_oid(src) + sizeof fid_ver(src));
784 dst->f_seq = be64_to_cpu(fid_seq(src));
785 dst->f_oid = be32_to_cpu(fid_oid(src));
786 dst->f_ver = be32_to_cpu(fid_ver(src));
787 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
790 static inline int fid_is_sane(const struct lu_fid *fid)
792 return fid != NULL &&
793 ((fid_seq(fid) >= FID_SEQ_START && fid_ver(fid) == 0) ||
794 fid_is_igif(fid) || fid_is_idif(fid) ||
795 fid_seq_is_rsvd(fid_seq(fid)));
798 static inline int fid_is_zero(const struct lu_fid *fid)
800 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
803 extern void lustre_swab_lu_fid(struct lu_fid *fid);
804 extern void lustre_swab_lu_seq_range(struct lu_seq_range *range);
806 static inline int lu_fid_eq(const struct lu_fid *f0,
807 const struct lu_fid *f1)
809 /* Check that there is no alignment padding. */
810 CLASSERT(sizeof *f0 ==
811 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
812 LASSERTF((fid_is_igif(f0) || fid_is_idif(f0)) ||
813 fid_ver(f0) == 0, DFID, PFID(f0));
814 LASSERTF((fid_is_igif(f1) || fid_is_idif(f1)) ||
815 fid_ver(f1) == 0, DFID, PFID(f1));
816 return memcmp(f0, f1, sizeof *f0) == 0;
819 #define __diff_normalize(val0, val1) \
821 typeof(val0) __val0 = (val0); \
822 typeof(val1) __val1 = (val1); \
824 (__val0 == __val1 ? 0 : __val0 > __val1 ? +1 : -1); \
827 static inline int lu_fid_cmp(const struct lu_fid *f0,
828 const struct lu_fid *f1)
831 __diff_normalize(fid_seq(f0), fid_seq(f1)) ?:
832 __diff_normalize(fid_oid(f0), fid_oid(f1)) ?:
833 __diff_normalize(fid_ver(f0), fid_ver(f1));
838 /** \defgroup lu_dir lu_dir
842 * Enumeration of possible directory entry attributes.
844 * Attributes follow directory entry header in the order they appear in this
847 enum lu_dirent_attrs {
850 LUDA_64BITHASH = 0x0004,
854 * Layout of readdir pages, as transmitted on wire.
857 /** valid if LUDA_FID is set. */
858 struct lu_fid lde_fid;
859 /** a unique entry identifier: a hash or an offset. */
861 /** total record length, including all attributes. */
865 /** optional variable size attributes following this entry.
866 * taken from enum lu_dirent_attrs.
869 /** name is followed by the attributes indicated in ->ldp_attrs, in
870 * their natural order. After the last attribute, padding bytes are
871 * added to make ->lde_reclen a multiple of 8.
877 * Definitions of optional directory entry attributes formats.
879 * Individual attributes do not have their length encoded in a generic way. It
880 * is assumed that consumer of an attribute knows its format. This means that
881 * it is impossible to skip over an unknown attribute, except by skipping over all
882 * remaining attributes (by using ->lde_reclen), which is not too
883 * constraining, because new server versions will append new attributes at
884 * the end of an entry.
888 * Fid directory attribute: a fid of an object referenced by the entry. This
889 * will be almost always requested by the client and supplied by the server.
891 * Aligned to 8 bytes.
893 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
898 * Aligned to 2 bytes.
905 __u64 ldp_hash_start;
909 struct lu_dirent ldp_entries[0];
912 enum lu_dirpage_flags {
914 * dirpage contains no entry.
918 * last entry's lde_hash equals ldp_hash_end.
923 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
925 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
928 return dp->ldp_entries;
931 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
933 struct lu_dirent *next;
935 if (le16_to_cpu(ent->lde_reclen) != 0)
936 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
943 static inline int lu_dirent_calc_size(int namelen, __u16 attr)
947 if (attr & LUDA_TYPE) {
948 const unsigned align = sizeof(struct luda_type) - 1;
949 size = (sizeof(struct lu_dirent) + namelen + align) & ~align;
950 size += sizeof(struct luda_type);
952 size = sizeof(struct lu_dirent) + namelen;
954 return (size + 7) & ~7;
957 static inline int lu_dirent_size(struct lu_dirent *ent)
959 if (le16_to_cpu(ent->lde_reclen) == 0) {
960 return lu_dirent_calc_size(le16_to_cpu(ent->lde_namelen),
961 le32_to_cpu(ent->lde_attrs));
963 return le16_to_cpu(ent->lde_reclen);
966 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
969 * MDS_READPAGE page size
971 * This is the directory page size packed in MDS_READPAGE RPC.
972 * It's different than CFS_PAGE_SIZE because the client needs to
973 * access the struct lu_dirpage header packed at the beginning of
974 * the "page" and without this there isn't any way to know find the
975 * lu_dirpage header is if client and server CFS_PAGE_SIZE differ.
977 #define LU_PAGE_SHIFT 12
978 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
979 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
981 #define LU_PAGE_COUNT (1 << (CFS_PAGE_SHIFT - LU_PAGE_SHIFT))
985 struct lustre_handle {
988 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
990 static inline int lustre_handle_is_used(struct lustre_handle *lh)
992 return lh->cookie != 0ull;
995 static inline int lustre_handle_equal(const struct lustre_handle *lh1,
996 const struct lustre_handle *lh2)
998 return lh1->cookie == lh2->cookie;
1001 static inline void lustre_handle_copy(struct lustre_handle *tgt,
1002 struct lustre_handle *src)
1004 tgt->cookie = src->cookie;
1007 /* flags for lm_flags */
1008 #define MSGHDR_AT_SUPPORT 0x1
1009 #define MSGHDR_CKSUM_INCOMPAT18 0x2
1011 #define lustre_msg lustre_msg_v2
1012 /* we depend on this structure to be 8-byte aligned */
1013 /* this type is only endian-adjusted in lustre_unpack_msg() */
1014 struct lustre_msg_v2 {
1023 __u32 lm_buflens[0];
1026 /* without gss, ptlrpc_body is put at the first buffer. */
1027 #define PTLRPC_NUM_VERSIONS 4
1028 #define JOBSTATS_JOBID_SIZE 32 /* 32 bytes string */
1029 struct ptlrpc_body_v3 {
1030 struct lustre_handle pb_handle;
1037 __u64 pb_last_committed;
1042 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1043 __u32 pb_service_time; /* for rep, actual service time */
1046 /* VBR: pre-versions */
1047 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1048 /* padding for future needs */
1049 __u64 pb_padding[4];
1050 char pb_jobid[JOBSTATS_JOBID_SIZE];
1052 #define ptlrpc_body ptlrpc_body_v3
1054 struct ptlrpc_body_v2 {
1055 struct lustre_handle pb_handle;
1062 __u64 pb_last_committed;
1067 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1068 __u32 pb_service_time; /* for rep, actual service time, also used for
1069 net_latency of req */
1072 /* VBR: pre-versions */
1073 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1074 /* padding for future needs */
1075 __u64 pb_padding[4];
1078 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
1080 /* message body offset for lustre_msg_v2 */
1081 /* ptlrpc body offset in all request/reply messages */
1082 #define MSG_PTLRPC_BODY_OFF 0
1084 /* normal request/reply message record offset */
1085 #define REQ_REC_OFF 1
1086 #define REPLY_REC_OFF 1
1088 /* ldlm request message body offset */
1089 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
1090 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
1092 /* ldlm intent lock message body offset */
1093 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
1094 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
1096 /* ldlm reply message body offset */
1097 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
1098 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
1100 /** only use in req->rq_{req,rep}_swab_mask */
1101 #define MSG_PTLRPC_HEADER_OFF 31
1103 /* Flags that are operation-specific go in the top 16 bits. */
1104 #define MSG_OP_FLAG_MASK 0xffff0000
1105 #define MSG_OP_FLAG_SHIFT 16
1107 /* Flags that apply to all requests are in the bottom 16 bits */
1108 #define MSG_GEN_FLAG_MASK 0x0000ffff
1109 #define MSG_LAST_REPLAY 0x0001
1110 #define MSG_RESENT 0x0002
1111 #define MSG_REPLAY 0x0004
1112 /* #define MSG_AT_SUPPORT 0x0008
1113 * This was used in early prototypes of adaptive timeouts, and while there
1114 * shouldn't be any users of that code there also isn't a need for using this
1115 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
1116 #define MSG_DELAY_REPLAY 0x0010
1117 #define MSG_VERSION_REPLAY 0x0020
1118 #define MSG_REQ_REPLAY_DONE 0x0040
1119 #define MSG_LOCK_REPLAY_DONE 0x0080
1122 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
1125 #define MSG_CONNECT_RECOVERING 0x00000001
1126 #define MSG_CONNECT_RECONNECT 0x00000002
1127 #define MSG_CONNECT_REPLAYABLE 0x00000004
1128 //#define MSG_CONNECT_PEER 0x8
1129 #define MSG_CONNECT_LIBCLIENT 0x00000010
1130 #define MSG_CONNECT_INITIAL 0x00000020
1131 #define MSG_CONNECT_ASYNC 0x00000040
1132 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
1133 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
1136 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
1137 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
1138 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
1139 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
1140 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
1141 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
1142 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
1143 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
1144 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
1145 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
1146 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
1147 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
1148 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
1149 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
1150 *We do not support JOIN FILE
1151 *anymore, reserve this flags
1152 *just for preventing such bit
1154 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
1155 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
1156 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /*Remote client */
1157 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /*Remote client by force */
1158 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
1159 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
1160 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
1161 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
1162 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
1163 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
1164 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
1165 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
1166 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
1167 #define OBD_CONNECT_REAL 0x8000000ULL /*real connection */
1168 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
1169 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
1170 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
1171 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
1172 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
1173 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
1174 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
1175 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
1176 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
1177 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
1178 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
1180 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
1181 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
1182 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
1183 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
1184 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
1185 * RPC error properly */
1186 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
1187 * finer space reservation */
1188 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
1189 * policy and 2.x server */
1190 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
1191 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
1192 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
1193 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
1195 * Please DO NOT add flag values here before first ensuring that this same
1196 * flag value is not in use on some other branch. Please clear any such
1197 * changes with senior engineers before starting to use a new flag. Then,
1198 * submit a small patch against EVERY branch that ONLY adds the new flag
1199 * and updates obd_connect_names[] for lprocfs_rd_connect_flags(), so it
1200 * can be approved and landed easily to reserve the flag for future use. */
1202 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
1203 * connection. It is a temporary bug fix for Imperative Recovery interop
1204 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
1205 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
1206 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
1208 #define OCD_HAS_FLAG(ocd, flg) \
1209 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
1212 #ifdef HAVE_LRU_RESIZE_SUPPORT
1213 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
1215 #define LRU_RESIZE_CONNECT_FLAG 0
1218 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
1219 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
1220 OBD_CONNECT_IBITS | \
1221 OBD_CONNECT_NODEVOH | OBD_CONNECT_ATTRFID | \
1222 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1223 OBD_CONNECT_RMT_CLIENT | \
1224 OBD_CONNECT_RMT_CLIENT_FORCE | \
1225 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_MDS_CAPA | \
1226 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_MDS_MDS | \
1227 OBD_CONNECT_FID | LRU_RESIZE_CONNECT_FLAG | \
1228 OBD_CONNECT_VBR | OBD_CONNECT_LOV_V3 | \
1229 OBD_CONNECT_SOM | OBD_CONNECT_FULL20 | \
1230 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
1231 OBD_CONNECT_EINPROGRESS | \
1232 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
1233 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK)
1234 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
1235 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
1236 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
1237 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_OSS_CAPA | \
1238 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1239 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_CKSUM | \
1240 OBD_CONNECT_RMT_CLIENT | \
1241 OBD_CONNECT_RMT_CLIENT_FORCE | OBD_CONNECT_VBR | \
1242 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
1243 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 | \
1244 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
1245 OBD_CONNECT_MAX_EASIZE | \
1246 OBD_CONNECT_EINPROGRESS | \
1247 OBD_CONNECT_JOBSTATS | \
1248 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
1249 OBD_CONNECT_LAYOUTLOCK)
1250 #define ECHO_CONNECT_SUPPORTED (0)
1251 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
1252 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
1253 OBD_CONNECT_MNE_SWAB)
1255 /* Features required for this version of the client to work with server */
1256 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
1259 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
1260 ((patch)<<8) + (fix))
1261 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
1262 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
1263 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
1264 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
1266 /* This structure is used for both request and reply.
1268 * If we eventually have separate connect data for different types, which we
1269 * almost certainly will, then perhaps we stick a union in here. */
1270 struct obd_connect_data_v1 {
1271 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
1272 __u32 ocd_version; /* lustre release version number */
1273 __u32 ocd_grant; /* initial cache grant amount (bytes) */
1274 __u32 ocd_index; /* LOV index to connect to */
1275 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
1276 __u64 ocd_ibits_known; /* inode bits this client understands */
1277 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
1278 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
1279 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
1280 __u32 ocd_unused; /* also fix lustre_swab_connect */
1281 __u64 ocd_transno; /* first transno from client to be replayed */
1282 __u32 ocd_group; /* MDS group on OST */
1283 __u32 ocd_cksum_types; /* supported checksum algorithms */
1284 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1285 __u32 ocd_instance; /* also fix lustre_swab_connect */
1286 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1289 struct obd_connect_data {
1290 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
1291 __u32 ocd_version; /* lustre release version number */
1292 __u32 ocd_grant; /* initial cache grant amount (bytes) */
1293 __u32 ocd_index; /* LOV index to connect to */
1294 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
1295 __u64 ocd_ibits_known; /* inode bits this client understands */
1296 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
1297 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
1298 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
1299 __u32 ocd_unused; /* also fix lustre_swab_connect */
1300 __u64 ocd_transno; /* first transno from client to be replayed */
1301 __u32 ocd_group; /* MDS group on OST */
1302 __u32 ocd_cksum_types; /* supported checksum algorithms */
1303 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1304 __u32 ocd_instance; /* instance # of this target */
1305 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1306 /* Fields after ocd_maxbytes are only accessible by the receiver
1307 * if the corresponding flag in ocd_connect_flags is set. Accessing
1308 * any field after ocd_maxbytes on the receiver without a valid flag
1309 * may result in out-of-bound memory access and kernel oops. */
1310 __u64 padding1; /* added 2.1.0. also fix lustre_swab_connect */
1311 __u64 padding2; /* added 2.1.0. also fix lustre_swab_connect */
1312 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
1313 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
1314 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
1315 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
1316 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
1317 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
1318 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
1319 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
1320 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
1321 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
1322 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
1323 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
1324 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1327 * Please DO NOT use any fields here before first ensuring that this same
1328 * field is not in use on some other branch. Please clear any such changes
1329 * with senior engineers before starting to use a new field. Then, submit
1330 * a small patch against EVERY branch that ONLY adds the new field along with
1331 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
1332 * reserve the flag for future use. */
1335 extern void lustre_swab_connect(struct obd_connect_data *ocd);
1338 * Supported checksum algorithms. Up to 32 checksum types are supported.
1339 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1340 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
1341 * algorithm and also the OBD_FL_CKSUM* flags.
1344 OBD_CKSUM_CRC32 = 0x00000001,
1345 OBD_CKSUM_ADLER = 0x00000002,
1346 OBD_CKSUM_CRC32C= 0x00000004,
1350 * OST requests: OBDO & OBD request records
1355 OST_REPLY = 0, /* reply ? */
1371 OST_QUOTACHECK = 18,
1373 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1376 #define OST_FIRST_OPC OST_REPLY
1379 OBD_FL_INLINEDATA = 0x00000001,
1380 OBD_FL_OBDMDEXISTS = 0x00000002,
1381 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1382 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1383 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1384 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1385 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1386 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1387 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1388 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1389 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1390 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1391 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1392 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1393 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
1394 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1395 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1396 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1397 * XXX: obsoleted - reserved for old
1398 * clients prior than 2.2 */
1399 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1400 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1402 /* Note that while these checksum values are currently separate bits,
1403 * in 2.x we can actually allow all values from 1-31 if we wanted. */
1404 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1405 OBD_FL_CKSUM_CRC32C,
1407 /* mask for local-only flag, which won't be sent over network */
1408 OBD_FL_LOCAL_MASK = 0xF0000000,
1411 #define LOV_MAGIC_V1 0x0BD10BD0
1412 #define LOV_MAGIC LOV_MAGIC_V1
1413 #define LOV_MAGIC_JOIN_V1 0x0BD20BD0
1414 #define LOV_MAGIC_V3 0x0BD30BD0
1417 * magic for fully defined striping
1418 * the idea is that we should have different magics for striping "hints"
1419 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1420 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1421 * we can't just change it w/o long way preparation, but we still need a
1422 * mechanism to allow LOD to differentiate hint versus ready striping.
1423 * so, at the moment we do a trick: MDT knows what to expect from request
1424 * depending on the case (replay uses ready striping, non-replay req uses
1425 * hints), so MDT replaces magic with appropriate one and now LOD can
1426 * easily understand what's inside -bzzz
1428 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
1429 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
1431 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
1432 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
1433 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
1434 #define LOV_PATTERN_CMOBD 0x200
1436 #define lov_ost_data lov_ost_data_v1
1437 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1438 __u64 l_object_id; /* OST object ID */
1439 __u64 l_object_seq; /* OST object seq number */
1440 __u32 l_ost_gen; /* generation of this l_ost_idx */
1441 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1444 #define lov_mds_md lov_mds_md_v1
1445 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1446 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1447 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1448 __u64 lmm_object_id; /* LOV object ID */
1449 __u64 lmm_object_seq; /* LOV object seq number */
1450 __u32 lmm_stripe_size; /* size of stripe in bytes */
1451 /* lmm_stripe_count used to be __u32 */
1452 __u16 lmm_stripe_count; /* num stripes in use for this object */
1453 __u16 lmm_layout_gen; /* layout generation number */
1454 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1457 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
1459 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1460 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1462 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1463 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1464 #define XATTR_USER_PREFIX "user."
1465 #define XATTR_TRUSTED_PREFIX "trusted."
1466 #define XATTR_SECURITY_PREFIX "security."
1467 #define XATTR_LUSTRE_PREFIX "lustre."
1469 #define XATTR_NAME_LOV "trusted.lov"
1470 #define XATTR_NAME_LMA "trusted.lma"
1471 #define XATTR_NAME_LMV "trusted.lmv"
1472 #define XATTR_NAME_LINK "trusted.link"
1473 #define XATTR_NAME_FID "trusted.fid"
1474 #define XATTR_NAME_VERSION "trusted.version"
1475 #define XATTR_NAME_SOM "trusted.som"
1476 #define XATTR_NAME_HSM "trusted.hsm"
1479 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1480 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1481 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1482 __u64 lmm_object_id; /* LOV object ID */
1483 __u64 lmm_object_seq; /* LOV object seq number */
1484 __u32 lmm_stripe_size; /* size of stripe in bytes */
1485 /* lmm_stripe_count used to be __u32 */
1486 __u16 lmm_stripe_count; /* num stripes in use for this object */
1487 __u16 lmm_layout_gen; /* layout generation number */
1488 char lmm_pool_name[LOV_MAXPOOLNAME]; /* must be 32bit aligned */
1489 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1493 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1494 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1495 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1496 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1497 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1498 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1499 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1500 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1501 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1502 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1503 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1504 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1505 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1506 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1507 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
1508 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1509 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1510 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1511 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1512 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1513 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1514 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1515 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
1516 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1517 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1518 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1519 /* ->mds if epoch opens or closes */
1520 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1521 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1522 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1523 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1524 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1526 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1527 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1528 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1530 /* OBD_MD_MDTIDX is used to get MDT index, but it is never been used overwire,
1531 * and it is already obsolete since 2.3 */
1532 /* #define OBD_MD_MDTIDX (0x0000000800000000ULL) */
1534 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1535 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1536 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1537 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1538 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
1539 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1540 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1541 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1542 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1543 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1545 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1547 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
1548 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
1549 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
1550 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
1552 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1554 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1555 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1556 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1557 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1558 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1560 /* don't forget obdo_fid which is way down at the bottom so it can
1561 * come after the definition of llog_cookie */
1564 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
1566 /* ost_body.data values for OST_BRW */
1568 #define OBD_BRW_READ 0x01
1569 #define OBD_BRW_WRITE 0x02
1570 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1571 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1572 * transfer and is not accounted in
1574 #define OBD_BRW_CHECK 0x10
1575 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1576 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1577 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1578 #define OBD_BRW_NOQUOTA 0x100
1579 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1580 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1581 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1582 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1583 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1585 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
1587 #define OST_MIN_PRECREATE 32
1588 #define OST_MAX_PRECREATE 20000
1597 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
1599 /* multiple of 8 bytes => can array */
1600 struct niobuf_remote {
1606 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
1608 /* lock value block communicated between the filter and llite */
1610 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1611 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1612 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1613 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1614 #define OST_LVB_IS_ERR(blocks) \
1615 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1616 #define OST_LVB_SET_ERR(blocks, rc) \
1617 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1618 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1628 extern void lustre_swab_ost_lvb_v1(struct ost_lvb_v1 *lvb);
1642 extern void lustre_swab_ost_lvb(struct ost_lvb *lvb);
1645 * lquota data structures
1648 #ifndef QUOTABLOCK_BITS
1649 #define QUOTABLOCK_BITS 10
1652 #ifndef QUOTABLOCK_SIZE
1653 #define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS)
1657 #define toqb(x) (((x) + QUOTABLOCK_SIZE - 1) >> QUOTABLOCK_BITS)
1660 /* The lquota_id structure is an union of all the possible identifier types that
1661 * can be used with quota, this includes:
1664 * - a FID which can be used for per-directory quota in the future */
1666 struct lu_fid qid_fid; /* FID for per-directory quota */
1667 __u64 qid_uid; /* user identifier */
1668 __u64 qid_gid; /* group identifier */
1671 /* quotactl management */
1672 struct obd_quotactl {
1674 __u32 qc_type; /* see Q_* flag below */
1677 struct obd_dqinfo qc_dqinfo;
1678 struct obd_dqblk qc_dqblk;
1681 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1683 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1684 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1685 #define Q_GETOINFO 0x800102 /* get obd quota info */
1686 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1687 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1689 #define Q_COPY(out, in, member) (out)->member = (in)->member
1691 #define QCTL_COPY(out, in) \
1693 Q_COPY(out, in, qc_cmd); \
1694 Q_COPY(out, in, qc_type); \
1695 Q_COPY(out, in, qc_id); \
1696 Q_COPY(out, in, qc_stat); \
1697 Q_COPY(out, in, qc_dqinfo); \
1698 Q_COPY(out, in, qc_dqblk); \
1701 /* Body of quota request used for quota acquire/release RPCs between quota
1702 * master (aka QMT) and slaves (ak QSD). */
1704 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1705 * and type (data or metadata) as well as
1706 * the quota type (user or group). */
1707 union lquota_id qb_id; /* uid or gid or directory FID */
1708 __u32 qb_flags; /* see below */
1710 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1711 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1712 __u64 qb_slv_ver; /* slave index file version */
1713 struct lustre_handle qb_lockh; /* per-ID lock handle */
1714 struct lustre_handle qb_glb_lockh; /* global lock handle */
1715 __u64 qb_padding1[4];
1718 /* When the quota_body is used in the reply of quota global intent
1719 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1720 #define qb_slv_fid qb_fid
1721 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1723 #define qb_qunit qb_usage
1725 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1726 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1727 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1728 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1730 extern void lustre_swab_quota_body(struct quota_body *b);
1732 /* Quota types currently supported */
1734 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1735 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1739 /* There are 2 different resource types on which a quota limit can be enforced:
1740 * - inodes on the MDTs
1741 * - blocks on the OSTs */
1743 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1744 LQUOTA_RES_DT = 0x02,
1746 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1748 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1751 * Space accounting support
1752 * Format of an accounting record, providing disk usage information for a given
1755 struct lquota_acct_rec { /* 16 bytes */
1756 __u64 bspace; /* current space in use */
1757 __u64 ispace; /* current # inodes in use */
1761 * Global quota index support
1762 * Format of a global record, providing global quota settings for a given quota
1765 struct lquota_glb_rec { /* 32 bytes */
1766 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1767 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1768 __u64 qbr_time; /* grace time, in seconds */
1769 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1774 * Slave index support
1775 * Format of a slave record, recording how much space is granted to a given
1778 struct lquota_slv_rec { /* 8 bytes */
1779 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1780 * in #inodes or kbytes */
1783 /* Data structures associated with the quota locks */
1785 /* Glimpse descriptor used for the index & per-ID quota locks */
1786 struct ldlm_gl_lquota_desc {
1787 union lquota_id gl_id; /* quota ID subject to the glimpse */
1788 __u64 gl_flags; /* see LQUOTA_FL* below */
1789 __u64 gl_ver; /* new index version */
1790 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1791 __u64 gl_softlimit; /* new softlimit */
1795 #define gl_qunit gl_hardlimit /* current qunit value used when
1796 * glimpsing per-ID quota locks */
1798 /* quota glimpse flags */
1799 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1801 /* LVB used with quota (global and per-ID) locks */
1803 __u64 lvb_flags; /* see LQUOTA_FL* above */
1804 __u64 lvb_id_may_rel; /* space that might be released later */
1805 __u64 lvb_id_rel; /* space released by the slave for this ID */
1806 __u64 lvb_id_qunit; /* current qunit value */
1810 extern void lustre_swab_lquota_lvb(struct lquota_lvb *lvb);
1812 /* LVB used with global quota lock */
1813 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1821 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1830 MDS_GETATTR_NAME = 34,
1835 MDS_DISCONNECT = 39,
1841 MDS_DONE_WRITING = 45,
1843 MDS_QUOTACHECK = 47,
1846 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1853 #define MDS_FIRST_OPC MDS_GETATTR
1870 } mds_reint_t, mdt_reint_t;
1872 extern void lustre_swab_generic_32s (__u32 *val);
1874 /* the disposition of the intent outlines what was executed */
1875 #define DISP_IT_EXECD 0x00000001
1876 #define DISP_LOOKUP_EXECD 0x00000002
1877 #define DISP_LOOKUP_NEG 0x00000004
1878 #define DISP_LOOKUP_POS 0x00000008
1879 #define DISP_OPEN_CREATE 0x00000010
1880 #define DISP_OPEN_OPEN 0x00000020
1881 #define DISP_ENQ_COMPLETE 0x00400000
1882 #define DISP_ENQ_OPEN_REF 0x00800000
1883 #define DISP_ENQ_CREATE_REF 0x01000000
1884 #define DISP_OPEN_LOCK 0x02000000
1886 /* INODE LOCK PARTS */
1887 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1888 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1889 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1890 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1892 #define MDS_INODELOCK_MAXSHIFT 3
1893 /* This FULL lock is useful to take on unlink sort of operations */
1894 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1896 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1898 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1899 * but was moved into name[1] along with the OID to avoid consuming the
1900 * name[2,3] fields that need to be used for the quota id (also a FID). */
1902 LUSTRE_RES_ID_SEQ_OFF = 0,
1903 LUSTRE_RES_ID_VER_OID_OFF = 1,
1904 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1905 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1906 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1907 LUSTRE_RES_ID_HSH_OFF = 3
1910 #define MDS_STATUS_CONN 1
1911 #define MDS_STATUS_LOV 2
1913 /* mdt_thread_info.mti_flags. */
1915 /* The flag indicates Size-on-MDS attributes are changed. */
1916 MF_SOM_CHANGE = (1 << 0),
1917 /* Flags indicates an epoch opens or closes. */
1918 MF_EPOCH_OPEN = (1 << 1),
1919 MF_EPOCH_CLOSE = (1 << 2),
1920 MF_MDC_CANCEL_FID1 = (1 << 3),
1921 MF_MDC_CANCEL_FID2 = (1 << 4),
1922 MF_MDC_CANCEL_FID3 = (1 << 5),
1923 MF_MDC_CANCEL_FID4 = (1 << 6),
1924 /* There is a pending attribute update. */
1925 MF_SOM_AU = (1 << 7),
1926 /* Cancel OST locks while getattr OST attributes. */
1927 MF_GETATTR_LOCK = (1 << 8),
1928 MF_GET_MDT_IDX = (1 << 9),
1931 #define MF_SOM_LOCAL_FLAGS (MF_SOM_CHANGE | MF_EPOCH_OPEN | MF_EPOCH_CLOSE)
1933 #define LUSTRE_BFLAG_UNCOMMITTED_WRITES 0x1
1935 /* these should be identical to their EXT4_*_FL counterparts, they are
1936 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1937 #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */
1938 #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */
1939 #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */
1940 #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */
1941 #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1944 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1945 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1946 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1947 * the S_* flags are kernel-internal values that change between kernel
1948 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1949 * See b=16526 for a full history. */
1950 static inline int ll_ext_to_inode_flags(int flags)
1952 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1953 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1954 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1955 #if defined(S_DIRSYNC)
1956 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1958 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0));
1961 static inline int ll_inode_to_ext_flags(int iflags)
1963 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1964 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1965 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1966 #if defined(S_DIRSYNC)
1967 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1969 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0));
1976 struct lustre_handle handle;
1978 __u64 size; /* Offset, in the case of MDS_READPAGE */
1982 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1991 __u32 flags; /* from vfs for pin/unpin, LUSTRE_BFLAG close */
1993 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1999 __u32 max_cookiesize;
2000 __u32 uid_h; /* high 32-bits of uid, for FUID */
2001 __u32 gid_h; /* high 32-bits of gid, for FUID */
2002 __u32 padding_5; /* also fix lustre_swab_mdt_body */
2010 extern void lustre_swab_mdt_body (struct mdt_body *b);
2012 struct mdt_ioepoch {
2013 struct lustre_handle handle;
2019 extern void lustre_swab_mdt_ioepoch (struct mdt_ioepoch *b);
2021 /* permissions for md_perm.mp_perm */
2023 CFS_SETUID_PERM = 0x01,
2024 CFS_SETGID_PERM = 0x02,
2025 CFS_SETGRP_PERM = 0x04,
2026 CFS_RMTACL_PERM = 0x08,
2027 CFS_RMTOWN_PERM = 0x10
2030 /* inode access permission for remote user, the inode info are omitted,
2031 * for client knows them. */
2032 struct mdt_remote_perm {
2039 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
2043 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
2045 struct mdt_rec_setattr {
2055 __u32 sa_padding_1_h;
2056 struct lu_fid sa_fid;
2065 __u32 sa_attr_flags;
2067 __u32 sa_bias; /* some operation flags */
2073 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
2076 * Attribute flags used in mdt_rec_setattr::sa_valid.
2077 * The kernel's #defines for ATTR_* should not be used over the network
2078 * since the client and MDS may run different kernels (see bug 13828)
2079 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
2081 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
2082 #define MDS_ATTR_UID 0x2ULL /* = 2 */
2083 #define MDS_ATTR_GID 0x4ULL /* = 4 */
2084 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
2085 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
2086 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
2087 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
2088 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
2089 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
2090 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
2091 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
2092 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
2093 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
2094 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
2095 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
2096 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
2099 #define FMODE_READ 00000001
2100 #define FMODE_WRITE 00000002
2103 #define MDS_FMODE_CLOSED 00000000
2104 #define MDS_FMODE_EXEC 00000004
2105 /* IO Epoch is opened on a closed file. */
2106 #define MDS_FMODE_EPOCH 01000000
2107 /* IO Epoch is opened on a file truncate. */
2108 #define MDS_FMODE_TRUNC 02000000
2109 /* Size-on-MDS Attribute Update is pending. */
2110 #define MDS_FMODE_SOM 04000000
2112 #define MDS_OPEN_CREATED 00000010
2113 #define MDS_OPEN_CROSS 00000020
2115 #define MDS_OPEN_CREAT 00000100
2116 #define MDS_OPEN_EXCL 00000200
2117 #define MDS_OPEN_TRUNC 00001000
2118 #define MDS_OPEN_APPEND 00002000
2119 #define MDS_OPEN_SYNC 00010000
2120 #define MDS_OPEN_DIRECTORY 00200000
2122 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
2123 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
2124 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
2125 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
2126 * We do not support JOIN FILE
2127 * anymore, reserve this flags
2128 * just for preventing such bit
2131 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
2132 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
2133 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
2134 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
2135 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
2138 /* permission for create non-directory file */
2139 #define MAY_CREATE (1 << 7)
2140 /* permission for create directory file */
2141 #define MAY_LINK (1 << 8)
2142 /* permission for delete from the directory */
2143 #define MAY_UNLINK (1 << 9)
2144 /* source's permission for rename */
2145 #define MAY_RENAME_SRC (1 << 10)
2146 /* target's permission for rename */
2147 #define MAY_RENAME_TAR (1 << 11)
2148 /* part (parent's) VTX permission check */
2149 #define MAY_VTX_PART (1 << 12)
2150 /* full VTX permission check */
2151 #define MAY_VTX_FULL (1 << 13)
2152 /* lfs rgetfacl permission check */
2153 #define MAY_RGETFACL (1 << 14)
2156 MDS_CHECK_SPLIT = 1 << 0,
2157 MDS_CROSS_REF = 1 << 1,
2158 MDS_VTX_BYPASS = 1 << 2,
2159 MDS_PERM_BYPASS = 1 << 3,
2161 MDS_QUOTA_IGNORE = 1 << 5,
2162 MDS_CLOSE_CLEANUP = 1 << 6,
2163 MDS_KEEP_ORPHAN = 1 << 7,
2164 MDS_RECOV_OPEN = 1 << 8,
2165 MDS_DATA_MODIFIED = 1 << 9,
2168 /* instance of mdt_reint_rec */
2169 struct mdt_rec_create {
2177 __u32 cr_suppgid1_h;
2179 __u32 cr_suppgid2_h;
2180 struct lu_fid cr_fid1;
2181 struct lu_fid cr_fid2;
2182 struct lustre_handle cr_old_handle; /* handle in case of open replay */
2186 __u64 cr_padding_1; /* rr_blocks */
2189 /* use of helpers set/get_mrc_cr_flags() is needed to access
2190 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2191 * extend cr_flags size without breaking 1.8 compat */
2192 __u32 cr_flags_l; /* for use with open, low 32 bits */
2193 __u32 cr_flags_h; /* for use with open, high 32 bits */
2194 __u32 cr_umask; /* umask for create */
2195 __u32 cr_padding_4; /* rr_padding_4 */
2198 static inline void set_mrc_cr_flags(struct mdt_rec_create *mrc, __u64 flags)
2200 mrc->cr_flags_l = (__u32)(flags & 0xFFFFFFFFUll);
2201 mrc->cr_flags_h = (__u32)(flags >> 32);
2204 static inline __u64 get_mrc_cr_flags(struct mdt_rec_create *mrc)
2206 return ((__u64)(mrc->cr_flags_l) | ((__u64)mrc->cr_flags_h << 32));
2209 /* instance of mdt_reint_rec */
2210 struct mdt_rec_link {
2218 __u32 lk_suppgid1_h;
2220 __u32 lk_suppgid2_h;
2221 struct lu_fid lk_fid1;
2222 struct lu_fid lk_fid2;
2224 __u64 lk_padding_1; /* rr_atime */
2225 __u64 lk_padding_2; /* rr_ctime */
2226 __u64 lk_padding_3; /* rr_size */
2227 __u64 lk_padding_4; /* rr_blocks */
2229 __u32 lk_padding_5; /* rr_mode */
2230 __u32 lk_padding_6; /* rr_flags */
2231 __u32 lk_padding_7; /* rr_padding_2 */
2232 __u32 lk_padding_8; /* rr_padding_3 */
2233 __u32 lk_padding_9; /* rr_padding_4 */
2236 /* instance of mdt_reint_rec */
2237 struct mdt_rec_unlink {
2245 __u32 ul_suppgid1_h;
2247 __u32 ul_suppgid2_h;
2248 struct lu_fid ul_fid1;
2249 struct lu_fid ul_fid2;
2251 __u64 ul_padding_2; /* rr_atime */
2252 __u64 ul_padding_3; /* rr_ctime */
2253 __u64 ul_padding_4; /* rr_size */
2254 __u64 ul_padding_5; /* rr_blocks */
2257 __u32 ul_padding_6; /* rr_flags */
2258 __u32 ul_padding_7; /* rr_padding_2 */
2259 __u32 ul_padding_8; /* rr_padding_3 */
2260 __u32 ul_padding_9; /* rr_padding_4 */
2263 /* instance of mdt_reint_rec */
2264 struct mdt_rec_rename {
2272 __u32 rn_suppgid1_h;
2274 __u32 rn_suppgid2_h;
2275 struct lu_fid rn_fid1;
2276 struct lu_fid rn_fid2;
2278 __u64 rn_padding_1; /* rr_atime */
2279 __u64 rn_padding_2; /* rr_ctime */
2280 __u64 rn_padding_3; /* rr_size */
2281 __u64 rn_padding_4; /* rr_blocks */
2282 __u32 rn_bias; /* some operation flags */
2283 __u32 rn_mode; /* cross-ref rename has mode */
2284 __u32 rn_padding_5; /* rr_flags */
2285 __u32 rn_padding_6; /* rr_padding_2 */
2286 __u32 rn_padding_7; /* rr_padding_3 */
2287 __u32 rn_padding_8; /* rr_padding_4 */
2290 /* instance of mdt_reint_rec */
2291 struct mdt_rec_setxattr {
2299 __u32 sx_suppgid1_h;
2301 __u32 sx_suppgid2_h;
2302 struct lu_fid sx_fid;
2303 __u64 sx_padding_1; /* These three are rr_fid2 */
2308 __u64 sx_padding_5; /* rr_ctime */
2309 __u64 sx_padding_6; /* rr_size */
2310 __u64 sx_padding_7; /* rr_blocks */
2313 __u32 sx_padding_8; /* rr_flags */
2314 __u32 sx_padding_9; /* rr_padding_2 */
2315 __u32 sx_padding_10; /* rr_padding_3 */
2316 __u32 sx_padding_11; /* rr_padding_4 */
2320 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2321 * Do NOT change the size of various members, otherwise the value
2322 * will be broken in lustre_swab_mdt_rec_reint().
2324 * If you add new members in other mdt_reint_xxx structres and need to use the
2325 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2327 struct mdt_rec_reint {
2335 __u32 rr_suppgid1_h;
2337 __u32 rr_suppgid2_h;
2338 struct lu_fid rr_fid1;
2339 struct lu_fid rr_fid2;
2348 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
2349 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
2350 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2353 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
2356 __u32 ld_tgt_count; /* how many MDS's */
2357 __u32 ld_active_tgt_count; /* how many active */
2358 __u32 ld_default_stripe_count; /* how many objects are used */
2359 __u32 ld_pattern; /* default MEA_MAGIC_* */
2360 __u64 ld_default_hash_size;
2361 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2362 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2363 __u32 ld_qos_maxage; /* in second */
2364 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2365 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2366 struct obd_uuid ld_uuid;
2369 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
2371 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
2372 struct lmv_stripe_md {
2377 char mea_pool_name[LOV_MAXPOOLNAME];
2378 struct lu_fid mea_ids[0];
2381 extern void lustre_swab_lmv_stripe_md(struct lmv_stripe_md *mea);
2383 /* lmv structures */
2384 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
2385 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
2386 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
2388 #define MAX_HASH_SIZE_32 0x7fffffffUL
2389 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
2390 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
2395 FLD_FIRST_OPC = FLD_QUERY
2401 SEQ_FIRST_OPC = SEQ_QUERY
2405 SEQ_ALLOC_SUPER = 0,
2410 * LOV data structures
2413 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2414 /* The size of the buffer the lov/mdc reserves for the
2415 * array of UUIDs returned by the MDS. With the current
2416 * protocol, this will limit the max number of OSTs per LOV */
2418 #define LOV_DESC_MAGIC 0xB0CCDE5C
2420 /* LOV settings descriptor (should only contain static info) */
2422 __u32 ld_tgt_count; /* how many OBD's */
2423 __u32 ld_active_tgt_count; /* how many active */
2424 __u32 ld_default_stripe_count; /* how many objects are used */
2425 __u32 ld_pattern; /* default PATTERN_RAID0 */
2426 __u64 ld_default_stripe_size; /* in bytes */
2427 __u64 ld_default_stripe_offset; /* in bytes */
2428 __u32 ld_padding_0; /* unused */
2429 __u32 ld_qos_maxage; /* in second */
2430 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2431 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2432 struct obd_uuid ld_uuid;
2435 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2437 extern void lustre_swab_lov_desc (struct lov_desc *ld);
2442 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2447 LDLM_BL_CALLBACK = 104,
2448 LDLM_CP_CALLBACK = 105,
2449 LDLM_GL_CALLBACK = 106,
2450 LDLM_SET_INFO = 107,
2453 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2455 #define RES_NAME_SIZE 4
2456 struct ldlm_res_id {
2457 __u64 name[RES_NAME_SIZE];
2460 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
2462 static inline int ldlm_res_eq(const struct ldlm_res_id *res0,
2463 const struct ldlm_res_id *res1)
2465 return !memcmp(res0, res1, sizeof(*res0));
2482 #define LCK_MODE_NUM 8
2492 #define LDLM_MIN_TYPE LDLM_PLAIN
2494 struct ldlm_extent {
2500 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
2501 struct ldlm_extent *ex2)
2503 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
2506 /* check if @ex1 contains @ex2 */
2507 static inline int ldlm_extent_contain(struct ldlm_extent *ex1,
2508 struct ldlm_extent *ex2)
2510 return (ex1->start <= ex2->start) && (ex1->end >= ex2->end);
2513 struct ldlm_inodebits {
2517 struct ldlm_flock_wire {
2525 /* it's important that the fields of the ldlm_extent structure match
2526 * the first fields of the ldlm_flock structure because there is only
2527 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2528 * this ever changes we will need to swab the union differently based
2529 * on the resource type. */
2532 struct ldlm_extent l_extent;
2533 struct ldlm_flock_wire l_flock;
2534 struct ldlm_inodebits l_inodebits;
2535 } ldlm_wire_policy_data_t;
2537 extern void lustre_swab_ldlm_policy_data (ldlm_wire_policy_data_t *d);
2539 union ldlm_gl_desc {
2540 struct ldlm_gl_lquota_desc lquota_desc;
2543 extern void lustre_swab_gl_desc(union ldlm_gl_desc *);
2545 struct ldlm_intent {
2549 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
2551 struct ldlm_resource_desc {
2552 ldlm_type_t lr_type;
2553 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
2554 struct ldlm_res_id lr_name;
2557 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
2559 struct ldlm_lock_desc {
2560 struct ldlm_resource_desc l_resource;
2561 ldlm_mode_t l_req_mode;
2562 ldlm_mode_t l_granted_mode;
2563 ldlm_wire_policy_data_t l_policy_data;
2566 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
2568 #define LDLM_LOCKREQ_HANDLES 2
2569 #define LDLM_ENQUEUE_CANCEL_OFF 1
2571 struct ldlm_request {
2574 struct ldlm_lock_desc lock_desc;
2575 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2578 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
2580 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
2581 * Otherwise, 2 are available. */
2582 #define ldlm_request_bufsize(count,type) \
2584 int _avail = LDLM_LOCKREQ_HANDLES; \
2585 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
2586 sizeof(struct ldlm_request) + \
2587 (count > _avail ? count - _avail : 0) * \
2588 sizeof(struct lustre_handle); \
2593 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2594 struct ldlm_lock_desc lock_desc;
2595 struct lustre_handle lock_handle;
2596 __u64 lock_policy_res1;
2597 __u64 lock_policy_res2;
2600 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
2602 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2603 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2606 * Opcodes for mountconf (mgs and mgc)
2611 MGS_EXCEPTION, /* node died, etc. */
2612 MGS_TARGET_REG, /* whenever target starts up */
2618 #define MGS_FIRST_OPC MGS_CONNECT
2620 #define MGS_PARAM_MAXLEN 1024
2621 #define KEY_SET_INFO "set_info"
2623 struct mgs_send_param {
2624 char mgs_param[MGS_PARAM_MAXLEN];
2627 /* We pass this info to the MGS so it can write config logs */
2628 #define MTI_NAME_MAXLEN 64
2629 #define MTI_PARAM_MAXLEN 4096
2630 #define MTI_NIDS_MAX 32
2631 struct mgs_target_info {
2632 __u32 mti_lustre_ver;
2633 __u32 mti_stripe_index;
2634 __u32 mti_config_ver;
2636 __u32 mti_nid_count;
2637 __u32 mti_instance; /* Running instance of target */
2638 char mti_fsname[MTI_NAME_MAXLEN];
2639 char mti_svname[MTI_NAME_MAXLEN];
2640 char mti_uuid[sizeof(struct obd_uuid)];
2641 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
2642 char mti_params[MTI_PARAM_MAXLEN];
2644 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
2646 struct mgs_nidtbl_entry {
2647 __u64 mne_version; /* table version of this entry */
2648 __u32 mne_instance; /* target instance # */
2649 __u32 mne_index; /* target index */
2650 __u32 mne_length; /* length of this entry - by bytes */
2651 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2652 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2653 __u8 mne_nid_size; /* size of each NID, by bytes */
2654 __u8 mne_nid_count; /* # of NIDs in buffer */
2656 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2659 extern void lustre_swab_mgs_nidtbl_entry(struct mgs_nidtbl_entry *oinfo);
2661 struct mgs_config_body {
2662 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2663 __u64 mcb_offset; /* next index of config log to request */
2664 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2666 __u8 mcb_bits; /* bits unit size of config log */
2667 __u32 mcb_units; /* # of units for bulk transfer */
2669 extern void lustre_swab_mgs_config_body(struct mgs_config_body *body);
2671 struct mgs_config_res {
2672 __u64 mcr_offset; /* index of last config log */
2673 __u64 mcr_size; /* size of the log */
2675 extern void lustre_swab_mgs_config_res(struct mgs_config_res *body);
2677 /* Config marker flags (in config log) */
2678 #define CM_START 0x01
2680 #define CM_SKIP 0x04
2681 #define CM_UPGRADE146 0x08
2682 #define CM_EXCLUDE 0x10
2683 #define CM_START_SKIP (CM_START | CM_SKIP)
2686 __u32 cm_step; /* aka config version */
2688 __u32 cm_vers; /* lustre release version number */
2689 __u32 cm_padding; /* 64 bit align */
2690 obd_time cm_createtime; /*when this record was first created */
2691 obd_time cm_canceltime; /*when this record is no longer valid*/
2692 char cm_tgtname[MTI_NAME_MAXLEN];
2693 char cm_comment[MTI_NAME_MAXLEN];
2696 extern void lustre_swab_cfg_marker(struct cfg_marker *marker,
2697 int swab, int size);
2700 * Opcodes for multiple servers.
2710 #define OBD_FIRST_OPC OBD_PING
2712 /* catalog of log objects */
2714 /** Identifier for a single log object */
2719 } __attribute__((packed));
2721 /** Records written to the CATALOGS list */
2722 #define CATLIST "CATALOGS"
2724 struct llog_logid lci_logid;
2728 } __attribute__((packed));
2730 /* Log data record types - there is no specific reason that these need to
2731 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2733 #define LLOG_OP_MAGIC 0x10600000
2734 #define LLOG_OP_MASK 0xfff00000
2737 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2738 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2739 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2740 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2741 REINT_UNLINK, /* obsolete after 2.5.0 */
2742 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2744 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2745 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2747 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2748 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2749 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2750 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2751 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2752 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2753 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2754 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2757 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2758 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2760 /** Log record header - stored in little endian order.
2761 * Each record must start with this struct, end with a llog_rec_tail,
2762 * and be a multiple of 256 bits in size.
2764 struct llog_rec_hdr {
2771 struct llog_rec_tail {
2776 struct llog_logid_rec {
2777 struct llog_rec_hdr lid_hdr;
2778 struct llog_logid lid_id;
2782 struct llog_rec_tail lid_tail;
2783 } __attribute__((packed));
2785 struct llog_unlink_rec {
2786 struct llog_rec_hdr lur_hdr;
2789 obd_count lur_count;
2790 struct llog_rec_tail lur_tail;
2791 } __attribute__((packed));
2793 struct llog_unlink64_rec {
2794 struct llog_rec_hdr lur_hdr;
2795 struct lu_fid lur_fid;
2796 obd_count lur_count; /* to destroy the lost precreated */
2800 struct llog_rec_tail lur_tail;
2801 } __attribute__((packed));
2803 struct llog_setattr64_rec {
2804 struct llog_rec_hdr lsr_hdr;
2812 struct llog_rec_tail lsr_tail;
2813 } __attribute__((packed));
2815 struct llog_size_change_rec {
2816 struct llog_rec_hdr lsc_hdr;
2817 struct ll_fid lsc_fid;
2822 struct llog_rec_tail lsc_tail;
2823 } __attribute__((packed));
2825 #define CHANGELOG_MAGIC 0xca103000
2827 /** \a changelog_rec_type's that can't be masked */
2828 #define CHANGELOG_MINMASK (1 << CL_MARK)
2829 /** bits covering all \a changelog_rec_type's */
2830 #define CHANGELOG_ALLMASK 0XFFFFFFFF
2831 /** default \a changelog_rec_type mask */
2832 #define CHANGELOG_DEFMASK CHANGELOG_ALLMASK & ~(1 << CL_ATIME | 1 << CL_CLOSE)
2834 /* changelog llog name, needed by client replicators */
2835 #define CHANGELOG_CATALOG "changelog_catalog"
2837 struct changelog_setinfo {
2840 } __attribute__((packed));
2842 /** changelog record */
2843 struct llog_changelog_rec {
2844 struct llog_rec_hdr cr_hdr;
2845 struct changelog_rec cr;
2846 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
2847 } __attribute__((packed));
2849 struct llog_changelog_ext_rec {
2850 struct llog_rec_hdr cr_hdr;
2851 struct changelog_ext_rec cr;
2852 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
2853 } __attribute__((packed));
2855 #define CHANGELOG_USER_PREFIX "cl"
2857 struct llog_changelog_user_rec {
2858 struct llog_rec_hdr cur_hdr;
2862 struct llog_rec_tail cur_tail;
2863 } __attribute__((packed));
2865 /* Old llog gen for compatibility */
2869 } __attribute__((packed));
2871 struct llog_gen_rec {
2872 struct llog_rec_hdr lgr_hdr;
2873 struct llog_gen lgr_gen;
2877 struct llog_rec_tail lgr_tail;
2880 /* On-disk header structure of each log object, stored in little endian order */
2881 #define LLOG_CHUNK_SIZE 8192
2882 #define LLOG_HEADER_SIZE (96)
2883 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2885 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2887 /* flags for the logs */
2889 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2890 LLOG_F_IS_CAT = 0x2,
2891 LLOG_F_IS_PLAIN = 0x4,
2894 struct llog_log_hdr {
2895 struct llog_rec_hdr llh_hdr;
2896 obd_time llh_timestamp;
2898 __u32 llh_bitmap_offset;
2902 /* for a catalog the first plain slot is next to it */
2903 struct obd_uuid llh_tgtuuid;
2904 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2905 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2906 struct llog_rec_tail llh_tail;
2907 } __attribute__((packed));
2909 #define LLOG_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2910 llh->llh_bitmap_offset - \
2911 sizeof(llh->llh_tail)) * 8)
2913 /** log cookies are used to reference a specific log file and a record therein */
2914 struct llog_cookie {
2915 struct llog_logid lgc_lgl;
2919 } __attribute__((packed));
2921 /** llog protocol */
2922 enum llogd_rpc_ops {
2923 LLOG_ORIGIN_HANDLE_CREATE = 501,
2924 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2925 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2926 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2927 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2928 LLOG_ORIGIN_CONNECT = 506,
2929 LLOG_CATINFO = 507, /* deprecated */
2930 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2931 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2933 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2937 struct llog_logid lgd_logid;
2939 __u32 lgd_llh_flags;
2941 __u32 lgd_saved_index;
2943 __u64 lgd_cur_offset;
2944 } __attribute__((packed));
2946 struct llogd_conn_body {
2947 struct llog_gen lgdc_gen;
2948 struct llog_logid lgdc_logid;
2949 __u32 lgdc_ctxt_idx;
2950 } __attribute__((packed));
2952 /* Note: 64-bit types are 64-bit aligned in structure */
2954 obd_valid o_valid; /* hot fields in this obdo */
2956 obd_id o_parent_seq;
2957 obd_size o_size; /* o_size-o_blocks == ost_lvb */
2961 obd_blocks o_blocks; /* brw: cli sent cached bytes */
2964 /* 32-bit fields start here: keep an even number of them via padding */
2965 obd_blksize o_blksize; /* optimal IO blocksize */
2966 obd_mode o_mode; /* brw: cli sent cache remain */
2970 obd_count o_nlink; /* brw: checksum */
2971 obd_count o_parent_oid;
2972 obd_count o_misc; /* brw: o_dropped */
2973 __u64 o_ioepoch; /* epoch in ost writes */
2974 __u32 o_stripe_idx; /* holds stripe idx */
2976 struct lustre_handle o_handle; /* brw: lock handle to prolong
2978 struct llog_cookie o_lcookie; /* destroy: unlink cookie from
2983 __u64 o_data_version; /* getattr: sum of iversion for
2985 * brw: grant space consumed on
2986 * the client for the write */
2992 #define o_id o_oi.oi_id
2993 #define o_seq o_oi.oi_seq
2994 #define o_dirty o_blocks
2995 #define o_undirty o_mode
2996 #define o_dropped o_misc
2997 #define o_cksum o_nlink
2998 #define o_grant_used o_data_version
3000 static inline void lustre_set_wire_obdo(struct obdo *wobdo, struct obdo *lobdo)
3002 memcpy(wobdo, lobdo, sizeof(*lobdo));
3003 wobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3006 static inline void lustre_get_wire_obdo(struct obdo *lobdo, struct obdo *wobdo)
3008 obd_flag local_flags = 0;
3010 if (lobdo->o_valid & OBD_MD_FLFLAGS)
3011 local_flags = lobdo->o_flags & OBD_FL_LOCAL_MASK;
3013 LASSERT(!(wobdo->o_flags & OBD_FL_LOCAL_MASK));
3015 memcpy(lobdo, wobdo, sizeof(*lobdo));
3016 if (local_flags != 0) {
3017 lobdo->o_valid |= OBD_MD_FLFLAGS;
3018 lobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3019 lobdo->o_flags |= local_flags;
3023 extern void lustre_swab_obdo (struct obdo *o);
3025 /* request structure for OST's */
3030 /* Key for FIEMAP to be used in get_info calls */
3031 struct ll_fiemap_info_key {
3034 struct ll_user_fiemap fiemap;
3037 extern void lustre_swab_ost_body (struct ost_body *b);
3038 extern void lustre_swab_ost_last_id(obd_id *id);
3039 extern void lustre_swab_fiemap(struct ll_user_fiemap *fiemap);
3041 extern void lustre_swab_lov_user_md_v1(struct lov_user_md_v1 *lum);
3042 extern void lustre_swab_lov_user_md_v3(struct lov_user_md_v3 *lum);
3043 extern void lustre_swab_lov_user_md_objects(struct lov_user_ost_data *lod,
3045 extern void lustre_swab_lov_mds_md(struct lov_mds_md *lmm);
3048 extern void lustre_swab_llogd_body (struct llogd_body *d);
3049 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
3050 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
3051 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec);
3054 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
3056 /* Functions for dumping PTLRPC fields */
3057 void dump_rniobuf(struct niobuf_remote *rnb);
3058 void dump_ioo(struct obd_ioobj *nb);
3059 void dump_obdo(struct obdo *oa);
3060 void dump_ost_body(struct ost_body *ob);
3061 void dump_rcs(__u32 *rc);
3063 #define IDX_INFO_MAGIC 0x3D37CC37
3065 /* Index file transfer through the network. The server serializes the index into
3066 * a byte stream which is sent to the client via a bulk transfer */
3070 /* reply: see idx_info_flags below */
3073 /* request & reply: number of lu_idxpage (to be) transferred */
3077 /* request: requested attributes passed down to the iterator API */
3080 /* request & reply: index file identifier (FID) */
3081 struct lu_fid ii_fid;
3083 /* reply: version of the index file before starting to walk the index.
3084 * Please note that the version can be modified at any time during the
3088 /* request: hash to start with:
3089 * reply: hash of the first entry of the first lu_idxpage and hash
3090 * of the entry to read next if any */
3091 __u64 ii_hash_start;
3094 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3098 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3106 extern void lustre_swab_idx_info(struct idx_info *ii);
3108 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3110 /* List of flags used in idx_info::ii_flags */
3111 enum idx_info_flags {
3112 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3113 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3114 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3115 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3118 #define LIP_MAGIC 0x8A6D6B6C
3120 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3122 /* 16-byte header */
3125 __u16 lip_nr; /* number of entries in the container */
3126 __u64 lip_pad0; /* additional padding for future use */
3128 /* key/record pairs are stored in the remaining 4080 bytes.
3129 * depending upon the flags in idx_info::ii_flags, each key/record
3130 * pair might be preceded by:
3132 * - the key size (II_FL_VARKEY is set)
3133 * - the record size (II_FL_VARREC is set)
3135 * For the time being, we only support fixed-size key & record. */
3136 char lip_entries[0];
3138 extern void lustre_swab_lip_header(struct lu_idxpage *lip);
3140 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3142 /* Gather all possible type associated with a 4KB container */
3144 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3145 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3146 char lp_array[LU_PAGE_SIZE];
3149 /* security opcodes */
3152 SEC_CTX_INIT_CONT = 802,
3155 SEC_FIRST_OPC = SEC_CTX_INIT
3159 * capa related definitions
3161 #define CAPA_HMAC_MAX_LEN 64
3162 #define CAPA_HMAC_KEY_MAX_LEN 56
3164 /* NB take care when changing the sequence of elements this struct,
3165 * because the offset info is used in find_capa() */
3166 struct lustre_capa {
3167 struct lu_fid lc_fid; /** fid */
3168 __u64 lc_opc; /** operations allowed */
3169 __u64 lc_uid; /** file owner */
3170 __u64 lc_gid; /** file group */
3171 __u32 lc_flags; /** HMAC algorithm & flags */
3172 __u32 lc_keyid; /** key# used for the capability */
3173 __u32 lc_timeout; /** capa timeout value (sec) */
3174 __u32 lc_expiry; /** expiry time (sec) */
3175 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
3176 } __attribute__((packed));
3178 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
3180 /** lustre_capa::lc_opc */
3182 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
3183 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
3184 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
3185 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
3186 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
3187 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
3188 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
3189 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
3190 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
3191 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
3192 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
3195 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3196 #define CAPA_OPC_MDS_ONLY \
3197 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3198 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3199 #define CAPA_OPC_OSS_ONLY \
3200 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3201 CAPA_OPC_OSS_DESTROY)
3202 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3203 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3205 /* MDS capability covers object capability for operations of body r/w
3206 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
3207 * while OSS capability only covers object capability for operations of
3208 * oss data(file content) r/w/truncate.
3210 static inline int capa_for_mds(struct lustre_capa *c)
3212 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
3215 static inline int capa_for_oss(struct lustre_capa *c)
3217 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
3220 /* lustre_capa::lc_hmac_alg */
3222 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
3226 #define CAPA_FL_MASK 0x00ffffff
3227 #define CAPA_HMAC_ALG_MASK 0xff000000
3229 struct lustre_capa_key {
3230 __u64 lk_seq; /**< mds# */
3231 __u32 lk_keyid; /**< key# */
3233 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
3234 } __attribute__((packed));
3236 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
3238 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3239 #define LINK_EA_MAGIC 0x11EAF1DFUL
3240 struct link_ea_header {
3243 __u64 leh_len; /* total size */
3249 /** Hardlink data is name and parent fid.
3250 * Stored in this crazy struct for maximum packing and endian-neutrality
3252 struct link_ea_entry {
3253 /** __u16 stored big-endian, unaligned */
3254 unsigned char lee_reclen[2];
3255 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3257 }__attribute__((packed));
3259 /** fid2path request/reply structure */
3260 struct getinfo_fid2path {
3261 struct lu_fid gf_fid;
3266 } __attribute__((packed));
3268 void lustre_swab_fid2path (struct getinfo_fid2path *gf);
3271 LAYOUT_INTENT_ACCESS = 0,
3272 LAYOUT_INTENT_READ = 1,
3273 LAYOUT_INTENT_WRITE = 2,
3274 LAYOUT_INTENT_GLIMPSE = 3,
3275 LAYOUT_INTENT_TRUNC = 4,
3276 LAYOUT_INTENT_RELEASE = 5,
3277 LAYOUT_INTENT_RESTORE = 6
3280 /* enqueue layout lock with intent */
3281 struct layout_intent {
3282 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3288 void lustre_swab_layout_intent(struct layout_intent *li);