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, 2013, 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>
101 #include <lustre/lustre_errno.h>
106 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
107 * FOO_REPLY_PORTAL is for incoming replies on the FOO
108 * FOO_BULK_PORTAL is for incoming bulk on the FOO
111 #define CONNMGR_REQUEST_PORTAL 1
112 #define CONNMGR_REPLY_PORTAL 2
113 //#define OSC_REQUEST_PORTAL 3
114 #define OSC_REPLY_PORTAL 4
115 //#define OSC_BULK_PORTAL 5
116 #define OST_IO_PORTAL 6
117 #define OST_CREATE_PORTAL 7
118 #define OST_BULK_PORTAL 8
119 //#define MDC_REQUEST_PORTAL 9
120 #define MDC_REPLY_PORTAL 10
121 //#define MDC_BULK_PORTAL 11
122 #define MDS_REQUEST_PORTAL 12
123 //#define MDS_REPLY_PORTAL 13
124 #define MDS_BULK_PORTAL 14
125 #define LDLM_CB_REQUEST_PORTAL 15
126 #define LDLM_CB_REPLY_PORTAL 16
127 #define LDLM_CANCEL_REQUEST_PORTAL 17
128 #define LDLM_CANCEL_REPLY_PORTAL 18
129 //#define PTLBD_REQUEST_PORTAL 19
130 //#define PTLBD_REPLY_PORTAL 20
131 //#define PTLBD_BULK_PORTAL 21
132 #define MDS_SETATTR_PORTAL 22
133 #define MDS_READPAGE_PORTAL 23
134 #define OUT_PORTAL 24
135 #define MGC_REPLY_PORTAL 25
136 #define MGS_REQUEST_PORTAL 26
137 #define MGS_REPLY_PORTAL 27
138 #define OST_REQUEST_PORTAL 28
139 #define FLD_REQUEST_PORTAL 29
140 #define SEQ_METADATA_PORTAL 30
141 #define SEQ_DATA_PORTAL 31
142 #define SEQ_CONTROLLER_PORTAL 32
143 #define MGS_BULK_PORTAL 33
145 /* Portal 63 is reserved for the Cray Inc DVS - nic@cray.com, roe@cray.com, n8851@cray.com */
148 #define PTL_RPC_MSG_REQUEST 4711
149 #define PTL_RPC_MSG_ERR 4712
150 #define PTL_RPC_MSG_REPLY 4713
152 /* DON'T use swabbed values of MAGIC as magic! */
153 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
154 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
156 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
157 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
159 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
161 #define PTLRPC_MSG_VERSION 0x00000003
162 #define LUSTRE_VERSION_MASK 0xffff0000
163 #define LUSTRE_OBD_VERSION 0x00010000
164 #define LUSTRE_MDS_VERSION 0x00020000
165 #define LUSTRE_OST_VERSION 0x00030000
166 #define LUSTRE_DLM_VERSION 0x00040000
167 #define LUSTRE_LOG_VERSION 0x00050000
168 #define LUSTRE_MGS_VERSION 0x00060000
170 typedef __u32 mdsno_t;
171 typedef __u64 seqno_t;
172 typedef __u64 obd_id;
173 typedef __u64 obd_seq;
174 typedef __s64 obd_time;
175 typedef __u64 obd_size;
176 typedef __u64 obd_off;
177 typedef __u64 obd_blocks;
178 typedef __u64 obd_valid;
179 typedef __u32 obd_blksize;
180 typedef __u32 obd_mode;
181 typedef __u32 obd_uid;
182 typedef __u32 obd_gid;
183 typedef __u32 obd_flag;
184 typedef __u32 obd_count;
187 * Describes a range of sequence, lsr_start is included but lsr_end is
189 * Same structure is used in fld module where lsr_index field holds mdt id
192 struct lu_seq_range {
199 #define LU_SEQ_RANGE_MDT 0x0
200 #define LU_SEQ_RANGE_OST 0x1
201 #define LU_SEQ_RANGE_ANY 0x3
203 #define LU_SEQ_RANGE_MASK 0x3
205 static inline unsigned fld_range_type(const struct lu_seq_range *range)
207 return range->lsr_flags & LU_SEQ_RANGE_MASK;
210 static inline int fld_range_is_ost(const struct lu_seq_range *range)
212 return fld_range_type(range) == LU_SEQ_RANGE_OST;
215 static inline int fld_range_is_mdt(const struct lu_seq_range *range)
217 return fld_range_type(range) == LU_SEQ_RANGE_MDT;
221 * This all range is only being used when fld client sends fld query request,
222 * but it does not know whether the seq is MDT or OST, so it will send req
223 * with ALL type, which means either seq type gotten from lookup can be
226 static inline unsigned fld_range_is_any(const struct lu_seq_range *range)
228 return fld_range_type(range) == LU_SEQ_RANGE_ANY;
231 static inline void fld_range_set_type(struct lu_seq_range *range,
234 LASSERT(!(flags & ~LU_SEQ_RANGE_MASK));
235 range->lsr_flags |= flags;
238 static inline void fld_range_set_mdt(struct lu_seq_range *range)
240 fld_range_set_type(range, LU_SEQ_RANGE_MDT);
243 static inline void fld_range_set_ost(struct lu_seq_range *range)
245 fld_range_set_type(range, LU_SEQ_RANGE_OST);
248 static inline void fld_range_set_any(struct lu_seq_range *range)
250 fld_range_set_type(range, LU_SEQ_RANGE_ANY);
254 * returns width of given range \a r
257 static inline __u64 range_space(const struct lu_seq_range *range)
259 return range->lsr_end - range->lsr_start;
263 * initialize range to zero
266 static inline void range_init(struct lu_seq_range *range)
268 range->lsr_start = range->lsr_end = range->lsr_index = 0;
272 * check if given seq id \a s is within given range \a r
275 static inline int range_within(const struct lu_seq_range *range,
278 return s >= range->lsr_start && s < range->lsr_end;
281 static inline int range_is_sane(const struct lu_seq_range *range)
283 return (range->lsr_end >= range->lsr_start);
286 static inline int range_is_zero(const struct lu_seq_range *range)
288 return (range->lsr_start == 0 && range->lsr_end == 0);
291 static inline int range_is_exhausted(const struct lu_seq_range *range)
294 return range_space(range) == 0;
297 /* return 0 if two range have the same location */
298 static inline int range_compare_loc(const struct lu_seq_range *r1,
299 const struct lu_seq_range *r2)
301 return r1->lsr_index != r2->lsr_index ||
302 r1->lsr_flags != r2->lsr_flags;
305 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x):%x:%s"
307 #define PRANGE(range) \
308 (range)->lsr_start, \
310 (range)->lsr_index, \
311 fld_range_is_mdt(range) ? "mdt" : "ost"
314 /** \defgroup lu_fid lu_fid
318 * Flags for lustre_mdt_attrs::lma_compat and lustre_mdt_attrs::lma_incompat.
319 * Deprecated since HSM and SOM attributes are now stored in separate on-disk
323 LMAC_HSM = 0x00000001,
324 LMAC_SOM = 0x00000002,
325 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
326 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
327 * under /O/<seq>/d<x>. */
331 * Masks for all features that should be supported by a Lustre version to
332 * access a specific file.
333 * This information is stored in lustre_mdt_attrs::lma_incompat.
336 LMAI_RELEASED = 0x00000001, /* file is released */
337 LMAI_AGENT = 0x00000002, /* agent inode */
338 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
339 is on the remote MDT */
341 #define LMA_INCOMPAT_SUPP (LMAI_AGENT | LMAI_REMOTE_PARENT)
343 extern void lustre_lma_swab(struct lustre_mdt_attrs *lma);
344 extern void lustre_lma_init(struct lustre_mdt_attrs *lma,
345 const struct lu_fid *fid,
346 __u32 compat, __u32 incompat);
348 * SOM on-disk attributes stored in a separate xattr.
351 /** Bitfield for supported data in this structure. For future use. */
354 /** Incompat feature list. The supported feature mask is availabe in
355 * SOM_INCOMPAT_SUPP */
358 /** IO Epoch SOM attributes belongs to */
360 /** total file size in objects */
362 /** total fs blocks in objects */
364 /** mds mount id the size is valid for */
367 extern void lustre_som_swab(struct som_attrs *attrs);
369 #define SOM_INCOMPAT_SUPP 0x0
372 * HSM on-disk attributes stored in a separate xattr.
375 /** Bitfield for supported data in this structure. For future use. */
378 /** HSM flags, see hsm_flags enum below */
380 /** backend archive id associated with the file */
382 /** version associated with the last archiving, if any */
385 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
391 /** LASTID file has zero OID */
392 LUSTRE_FID_LASTID_OID = 0UL,
393 /** initial fid id value */
394 LUSTRE_FID_INIT_OID = 1UL
397 /** returns fid object sequence */
398 static inline __u64 fid_seq(const struct lu_fid *fid)
403 /** returns fid object id */
404 static inline __u32 fid_oid(const struct lu_fid *fid)
409 /** returns fid object version */
410 static inline __u32 fid_ver(const struct lu_fid *fid)
415 static inline void fid_zero(struct lu_fid *fid)
417 memset(fid, 0, sizeof(*fid));
420 static inline obd_id fid_ver_oid(const struct lu_fid *fid)
422 return ((__u64)fid_ver(fid) << 32 | fid_oid(fid));
426 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
427 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
428 * used for other purposes and not risk collisions with existing inodes.
430 * Different FID Format
431 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
434 FID_SEQ_OST_MDT0 = 0,
435 FID_SEQ_LLOG = 1, /* unnamed llogs */
437 FID_SEQ_OST_MDT1 = 3,
438 FID_SEQ_OST_MAX = 9, /* Max MDT count before OST_on_FID */
439 FID_SEQ_LLOG_NAME = 10, /* named llogs */
442 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
443 FID_SEQ_IDIF = 0x100000000ULL,
444 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
445 /* Normal FID sequence starts from this value, i.e. 1<<33 */
446 FID_SEQ_START = 0x200000000ULL,
447 /* sequence for local pre-defined FIDs listed in local_oid */
448 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
449 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
450 /* sequence is used for local named objects FIDs generated
451 * by local_object_storage library */
452 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
453 /* Because current FLD will only cache the fid sequence, instead
454 * of oid on the client side, if the FID needs to be exposed to
455 * clients sides, it needs to make sure all of fids under one
456 * sequence will be located in one MDT. */
457 FID_SEQ_SPECIAL = 0x200000004ULL,
458 FID_SEQ_QUOTA = 0x200000005ULL,
459 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
460 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
461 FID_SEQ_NORMAL = 0x200000400ULL,
462 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
465 #define OBIF_OID_MAX_BITS 32
466 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
467 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
468 #define IDIF_OID_MAX_BITS 48
469 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
470 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
472 /** OID for FID_SEQ_SPECIAL */
474 /* Big Filesystem Lock to serialize rename operations */
475 FID_OID_SPECIAL_BFL = 1UL,
478 /** OID for FID_SEQ_DOT_LUSTRE */
479 enum dot_lustre_oid {
480 FID_OID_DOT_LUSTRE = 1UL,
481 FID_OID_DOT_LUSTRE_OBF = 2UL,
484 static inline int fid_seq_is_mdt0(obd_seq seq)
486 return (seq == FID_SEQ_OST_MDT0);
489 static inline int fid_seq_is_mdt(const __u64 seq)
491 return seq == FID_SEQ_OST_MDT0 || seq >= FID_SEQ_NORMAL;
494 static inline int fid_seq_is_echo(obd_seq seq)
496 return (seq == FID_SEQ_ECHO);
499 static inline int fid_is_echo(const struct lu_fid *fid)
501 return fid_seq_is_echo(fid_seq(fid));
504 static inline int fid_seq_is_llog(obd_seq seq)
506 return (seq == FID_SEQ_LLOG);
509 static inline int fid_is_llog(const struct lu_fid *fid)
511 /* file with OID == 0 is not llog but contains last oid */
512 return fid_seq_is_llog(fid_seq(fid)) && fid_oid(fid) > 0;
515 static inline int fid_seq_is_rsvd(const __u64 seq)
517 return (seq > FID_SEQ_OST_MDT0 && seq <= FID_SEQ_RSVD);
520 static inline int fid_seq_is_special(const __u64 seq)
522 return seq == FID_SEQ_SPECIAL;
525 static inline int fid_seq_is_local_file(const __u64 seq)
527 return seq == FID_SEQ_LOCAL_FILE ||
528 seq == FID_SEQ_LOCAL_NAME;
531 static inline int fid_seq_is_root(const __u64 seq)
533 return seq == FID_SEQ_ROOT;
536 static inline int fid_seq_is_dot(const __u64 seq)
538 return seq == FID_SEQ_DOT_LUSTRE;
541 static inline int fid_seq_is_default(const __u64 seq)
543 return seq == FID_SEQ_LOV_DEFAULT;
546 static inline int fid_is_mdt0(const struct lu_fid *fid)
548 return fid_seq_is_mdt0(fid_seq(fid));
551 static inline void lu_root_fid(struct lu_fid *fid)
553 fid->f_seq = FID_SEQ_ROOT;
559 * Check if a fid is igif or not.
560 * \param fid the fid to be tested.
561 * \return true if the fid is a igif; otherwise false.
563 static inline int fid_seq_is_igif(const __u64 seq)
565 return seq >= FID_SEQ_IGIF && seq <= FID_SEQ_IGIF_MAX;
568 static inline int fid_is_igif(const struct lu_fid *fid)
570 return fid_seq_is_igif(fid_seq(fid));
574 * Check if a fid is idif or not.
575 * \param fid the fid to be tested.
576 * \return true if the fid is a idif; otherwise false.
578 static inline int fid_seq_is_idif(const __u64 seq)
580 return seq >= FID_SEQ_IDIF && seq <= FID_SEQ_IDIF_MAX;
583 static inline int fid_is_idif(const struct lu_fid *fid)
585 return fid_seq_is_idif(fid_seq(fid));
588 static inline int fid_is_local_file(const struct lu_fid *fid)
590 return fid_seq_is_local_file(fid_seq(fid));
593 static inline int fid_seq_is_norm(const __u64 seq)
595 return (seq >= FID_SEQ_NORMAL);
598 static inline int fid_is_norm(const struct lu_fid *fid)
600 return fid_seq_is_norm(fid_seq(fid));
603 /* convert an OST objid into an IDIF FID SEQ number */
604 static inline obd_seq fid_idif_seq(obd_id id, __u32 ost_idx)
606 return FID_SEQ_IDIF | (ost_idx << 16) | ((id >> 32) & 0xffff);
609 /* convert a packed IDIF FID into an OST objid */
610 static inline obd_id fid_idif_id(obd_seq seq, __u32 oid, __u32 ver)
612 return ((__u64)ver << 48) | ((seq & 0xffff) << 32) | oid;
615 /* extract ost index from IDIF FID */
616 static inline __u32 fid_idif_ost_idx(const struct lu_fid *fid)
618 LASSERT(fid_is_idif(fid));
619 return (fid_seq(fid) >> 16) & 0xffff;
622 /* extract OST sequence (group) from a wire ost_id (id/seq) pair */
623 static inline obd_seq ostid_seq(const struct ost_id *ostid)
625 if (fid_seq_is_mdt0(ostid->oi.oi_seq))
626 return FID_SEQ_OST_MDT0;
628 if (fid_seq_is_default(ostid->oi.oi_seq))
629 return FID_SEQ_LOV_DEFAULT;
631 if (fid_is_idif(&ostid->oi_fid))
632 return FID_SEQ_OST_MDT0;
634 return fid_seq(&ostid->oi_fid);
637 /* extract OST objid from a wire ost_id (id/seq) pair */
638 static inline obd_id ostid_id(const struct ost_id *ostid)
640 if (fid_seq_is_mdt0(ostid_seq(ostid)))
641 return ostid->oi.oi_id & IDIF_OID_MASK;
643 if (fid_is_idif(&ostid->oi_fid))
644 return fid_idif_id(fid_seq(&ostid->oi_fid),
645 fid_oid(&ostid->oi_fid), 0);
647 return fid_oid(&ostid->oi_fid);
650 static inline void ostid_set_seq(struct ost_id *oi, __u64 seq)
652 if (fid_seq_is_mdt0(seq) || fid_seq_is_default(seq)) {
655 oi->oi_fid.f_seq = seq;
656 /* Note: if f_oid + f_ver is zero, we need init it
657 * to be 1, otherwise, ostid_seq will treat this
658 * as old ostid (oi_seq == 0) */
659 if (oi->oi_fid.f_oid == 0 && oi->oi_fid.f_ver == 0)
660 oi->oi_fid.f_oid = LUSTRE_FID_INIT_OID;
664 static inline void ostid_set_seq_mdt0(struct ost_id *oi)
666 ostid_set_seq(oi, FID_SEQ_OST_MDT0);
669 static inline void ostid_set_seq_echo(struct ost_id *oi)
671 ostid_set_seq(oi, FID_SEQ_ECHO);
674 static inline void ostid_set_seq_llog(struct ost_id *oi)
676 ostid_set_seq(oi, FID_SEQ_LLOG);
680 * Note: we need check oi_seq to decide where to set oi_id,
681 * so oi_seq should always be set ahead of oi_id.
683 static inline void ostid_set_id(struct ost_id *oi, __u64 oid)
685 if (fid_seq_is_mdt0(ostid_seq(oi))) {
686 if (oid >= IDIF_MAX_OID) {
687 CERROR("Bad "LPU64" to set "DOSTID"\n",
693 if (oid > OBIF_MAX_OID) {
694 CERROR("Bad "LPU64" to set "DOSTID"\n",
698 oi->oi_fid.f_oid = oid;
702 static inline void ostid_inc_id(struct ost_id *oi)
704 if (fid_seq_is_mdt0(ostid_seq(oi))) {
705 if (unlikely(ostid_id(oi) + 1 > IDIF_MAX_OID)) {
706 CERROR("Bad inc "DOSTID"\n", POSTID(oi));
715 static inline void ostid_dec_id(struct ost_id *oi)
717 if (fid_seq_is_mdt0(ostid_seq(oi)))
724 * Unpack an OST object id/seq (group) into a FID. This is needed for
725 * converting all obdo, lmm, lsm, etc. 64-bit id/seq pairs into proper
726 * FIDs. Note that if an id/seq is already in FID/IDIF format it will
727 * be passed through unchanged. Only legacy OST objects in "group 0"
728 * will be mapped into the IDIF namespace so that they can fit into the
729 * struct lu_fid fields without loss. For reference see:
730 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs
732 static inline int ostid_to_fid(struct lu_fid *fid, struct ost_id *ostid,
735 if (ost_idx > 0xffff) {
736 CERROR("bad ost_idx, "DOSTID" ost_idx:%u\n", POSTID(ostid),
741 if (fid_seq_is_mdt0(ostid_seq(ostid))) {
742 /* This is a "legacy" (old 1.x/2.early) OST object in "group 0"
743 * that we map into the IDIF namespace. It allows up to 2^48
744 * objects per OST, as this is the object namespace that has
745 * been in production for years. This can handle create rates
746 * of 1M objects/s/OST for 9 years, or combinations thereof. */
747 if (ostid_id(ostid) >= IDIF_MAX_OID) {
748 CERROR("bad MDT0 id, "DOSTID" ost_idx:%u\n",
749 POSTID(ostid), ost_idx);
752 fid->f_seq = fid_idif_seq(ostid_id(ostid), ost_idx);
753 /* truncate to 32 bits by assignment */
754 fid->f_oid = ostid_id(ostid);
755 /* in theory, not currently used */
756 fid->f_ver = ostid_id(ostid) >> 48;
757 } else /* if (fid_seq_is_idif(seq) || fid_seq_is_norm(seq)) */ {
758 /* This is either an IDIF object, which identifies objects across
759 * all OSTs, or a regular FID. The IDIF namespace maps legacy
760 * OST objects into the FID namespace. In both cases, we just
761 * pass the FID through, no conversion needed. */
762 if (ostid->oi_fid.f_ver != 0) {
763 CERROR("bad MDT0 id, "DOSTID" ost_idx:%u\n",
764 POSTID(ostid), ost_idx);
767 *fid = ostid->oi_fid;
773 /* pack any OST FID into an ostid (id/seq) for the wire/disk */
774 static inline int fid_to_ostid(const struct lu_fid *fid, struct ost_id *ostid)
776 if (unlikely(fid_seq_is_igif(fid->f_seq))) {
777 CERROR("bad IGIF, "DFID"\n", PFID(fid));
781 if (fid_is_idif(fid)) {
782 ostid_set_seq_mdt0(ostid);
783 ostid_set_id(ostid, fid_idif_id(fid_seq(fid), fid_oid(fid),
786 ostid->oi_fid = *fid;
792 /* Check whether the fid is for LAST_ID */
793 static inline int fid_is_last_id(const struct lu_fid *fid)
795 return (fid_oid(fid) == 0);
799 * Get inode number from a igif.
800 * \param fid a igif to get inode number from.
801 * \return inode number for the igif.
803 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
808 extern void lustre_swab_ost_id(struct ost_id *oid);
811 * Get inode generation from a igif.
812 * \param fid a igif to get inode generation from.
813 * \return inode generation for the igif.
815 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
821 * Build igif from the inode number/generation.
823 static inline void lu_igif_build(struct lu_fid *fid, __u32 ino, __u32 gen)
831 * Fids are transmitted across network (in the sender byte-ordering),
832 * and stored on disk in big-endian order.
834 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
836 /* check that all fields are converted */
837 CLASSERT(sizeof *src ==
838 sizeof fid_seq(src) +
839 sizeof fid_oid(src) + sizeof fid_ver(src));
840 dst->f_seq = cpu_to_le64(fid_seq(src));
841 dst->f_oid = cpu_to_le32(fid_oid(src));
842 dst->f_ver = cpu_to_le32(fid_ver(src));
845 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
847 /* check that all fields are converted */
848 CLASSERT(sizeof *src ==
849 sizeof fid_seq(src) +
850 sizeof fid_oid(src) + sizeof fid_ver(src));
851 dst->f_seq = le64_to_cpu(fid_seq(src));
852 dst->f_oid = le32_to_cpu(fid_oid(src));
853 dst->f_ver = le32_to_cpu(fid_ver(src));
856 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
858 /* check that all fields are converted */
859 CLASSERT(sizeof *src ==
860 sizeof fid_seq(src) +
861 sizeof fid_oid(src) + sizeof fid_ver(src));
862 dst->f_seq = cpu_to_be64(fid_seq(src));
863 dst->f_oid = cpu_to_be32(fid_oid(src));
864 dst->f_ver = cpu_to_be32(fid_ver(src));
867 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
869 /* check that all fields are converted */
870 CLASSERT(sizeof *src ==
871 sizeof fid_seq(src) +
872 sizeof fid_oid(src) + sizeof fid_ver(src));
873 dst->f_seq = be64_to_cpu(fid_seq(src));
874 dst->f_oid = be32_to_cpu(fid_oid(src));
875 dst->f_ver = be32_to_cpu(fid_ver(src));
878 static inline int fid_is_sane(const struct lu_fid *fid)
880 return fid != NULL &&
881 ((fid_seq(fid) >= FID_SEQ_START && fid_ver(fid) == 0) ||
882 fid_is_igif(fid) || fid_is_idif(fid) ||
883 fid_seq_is_rsvd(fid_seq(fid)));
886 static inline int fid_is_zero(const struct lu_fid *fid)
888 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
891 extern void lustre_swab_lu_fid(struct lu_fid *fid);
892 extern void lustre_swab_lu_seq_range(struct lu_seq_range *range);
894 static inline int lu_fid_eq(const struct lu_fid *f0, const struct lu_fid *f1)
896 /* Check that there is no alignment padding. */
897 CLASSERT(sizeof *f0 ==
898 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
899 return memcmp(f0, f1, sizeof *f0) == 0;
902 #define __diff_normalize(val0, val1) \
904 typeof(val0) __val0 = (val0); \
905 typeof(val1) __val1 = (val1); \
907 (__val0 == __val1 ? 0 : __val0 > __val1 ? +1 : -1); \
910 static inline int lu_fid_cmp(const struct lu_fid *f0,
911 const struct lu_fid *f1)
914 __diff_normalize(fid_seq(f0), fid_seq(f1)) ?:
915 __diff_normalize(fid_oid(f0), fid_oid(f1)) ?:
916 __diff_normalize(fid_ver(f0), fid_ver(f1));
919 static inline void ostid_cpu_to_le(const struct ost_id *src_oi,
920 struct ost_id *dst_oi)
922 if (fid_seq_is_mdt0(ostid_seq(src_oi))) {
923 dst_oi->oi.oi_id = cpu_to_le64(src_oi->oi.oi_id);
924 dst_oi->oi.oi_seq = cpu_to_le64(src_oi->oi.oi_seq);
926 fid_cpu_to_le(&dst_oi->oi_fid, &src_oi->oi_fid);
930 static inline void ostid_le_to_cpu(const struct ost_id *src_oi,
931 struct ost_id *dst_oi)
933 if (fid_seq_is_mdt0(ostid_seq(src_oi))) {
934 dst_oi->oi.oi_id = le64_to_cpu(src_oi->oi.oi_id);
935 dst_oi->oi.oi_seq = le64_to_cpu(src_oi->oi.oi_seq);
937 fid_le_to_cpu(&dst_oi->oi_fid, &src_oi->oi_fid);
943 /** \defgroup lu_dir lu_dir
947 * Enumeration of possible directory entry attributes.
949 * Attributes follow directory entry header in the order they appear in this
952 enum lu_dirent_attrs {
955 LUDA_64BITHASH = 0x0004,
957 /* The following attrs are used for MDT interanl only,
958 * not visible to client */
960 /* Verify the dirent consistency */
961 LUDA_VERIFY = 0x8000,
962 /* Only check but not repair the dirent inconsistency */
963 LUDA_VERIFY_DRYRUN = 0x4000,
964 /* The dirent has been repaired, or to be repaired (dryrun). */
965 LUDA_REPAIR = 0x2000,
966 /* The system is upgraded, has beed or to be repaired (dryrun). */
967 LUDA_UPGRADE = 0x1000,
968 /* Ignore this record, go to next directly. */
969 LUDA_IGNORE = 0x0800,
972 #define LU_DIRENT_ATTRS_MASK 0xf800
975 * Layout of readdir pages, as transmitted on wire.
978 /** valid if LUDA_FID is set. */
979 struct lu_fid lde_fid;
980 /** a unique entry identifier: a hash or an offset. */
982 /** total record length, including all attributes. */
986 /** optional variable size attributes following this entry.
987 * taken from enum lu_dirent_attrs.
990 /** name is followed by the attributes indicated in ->ldp_attrs, in
991 * their natural order. After the last attribute, padding bytes are
992 * added to make ->lde_reclen a multiple of 8.
998 * Definitions of optional directory entry attributes formats.
1000 * Individual attributes do not have their length encoded in a generic way. It
1001 * is assumed that consumer of an attribute knows its format. This means that
1002 * it is impossible to skip over an unknown attribute, except by skipping over all
1003 * remaining attributes (by using ->lde_reclen), which is not too
1004 * constraining, because new server versions will append new attributes at
1005 * the end of an entry.
1009 * Fid directory attribute: a fid of an object referenced by the entry. This
1010 * will be almost always requested by the client and supplied by the server.
1012 * Aligned to 8 bytes.
1014 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
1019 * Aligned to 2 bytes.
1026 __u64 ldp_hash_start;
1030 struct lu_dirent ldp_entries[0];
1033 enum lu_dirpage_flags {
1035 * dirpage contains no entry.
1039 * last entry's lde_hash equals ldp_hash_end.
1041 LDF_COLLIDE = 1 << 1
1044 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
1046 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
1049 return dp->ldp_entries;
1052 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
1054 struct lu_dirent *next;
1056 if (le16_to_cpu(ent->lde_reclen) != 0)
1057 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
1064 static inline int lu_dirent_calc_size(int namelen, __u16 attr)
1068 if (attr & LUDA_TYPE) {
1069 const unsigned align = sizeof(struct luda_type) - 1;
1070 size = (sizeof(struct lu_dirent) + namelen + align) & ~align;
1071 size += sizeof(struct luda_type);
1073 size = sizeof(struct lu_dirent) + namelen;
1075 return (size + 7) & ~7;
1078 static inline int lu_dirent_size(struct lu_dirent *ent)
1080 if (le16_to_cpu(ent->lde_reclen) == 0) {
1081 return lu_dirent_calc_size(le16_to_cpu(ent->lde_namelen),
1082 le32_to_cpu(ent->lde_attrs));
1084 return le16_to_cpu(ent->lde_reclen);
1087 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
1090 * MDS_READPAGE page size
1092 * This is the directory page size packed in MDS_READPAGE RPC.
1093 * It's different than PAGE_CACHE_SIZE because the client needs to
1094 * access the struct lu_dirpage header packed at the beginning of
1095 * the "page" and without this there isn't any way to know find the
1096 * lu_dirpage header is if client and server PAGE_CACHE_SIZE differ.
1098 #define LU_PAGE_SHIFT 12
1099 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
1100 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
1102 #define LU_PAGE_COUNT (1 << (PAGE_CACHE_SHIFT - LU_PAGE_SHIFT))
1106 struct lustre_handle {
1109 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
1111 static inline int lustre_handle_is_used(struct lustre_handle *lh)
1113 return lh->cookie != 0ull;
1116 static inline int lustre_handle_equal(const struct lustre_handle *lh1,
1117 const struct lustre_handle *lh2)
1119 return lh1->cookie == lh2->cookie;
1122 static inline void lustre_handle_copy(struct lustre_handle *tgt,
1123 struct lustre_handle *src)
1125 tgt->cookie = src->cookie;
1128 /* flags for lm_flags */
1129 #define MSGHDR_AT_SUPPORT 0x1
1130 #define MSGHDR_CKSUM_INCOMPAT18 0x2
1132 #define lustre_msg lustre_msg_v2
1133 /* we depend on this structure to be 8-byte aligned */
1134 /* this type is only endian-adjusted in lustre_unpack_msg() */
1135 struct lustre_msg_v2 {
1144 __u32 lm_buflens[0];
1147 /* without gss, ptlrpc_body is put at the first buffer. */
1148 #define PTLRPC_NUM_VERSIONS 4
1149 #define JOBSTATS_JOBID_SIZE 32 /* 32 bytes string */
1150 struct ptlrpc_body_v3 {
1151 struct lustre_handle pb_handle;
1158 __u64 pb_last_committed;
1163 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1164 __u32 pb_service_time; /* for rep, actual service time */
1167 /* VBR: pre-versions */
1168 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1169 /* padding for future needs */
1170 __u64 pb_padding[4];
1171 char pb_jobid[JOBSTATS_JOBID_SIZE];
1173 #define ptlrpc_body ptlrpc_body_v3
1175 struct ptlrpc_body_v2 {
1176 struct lustre_handle pb_handle;
1183 __u64 pb_last_committed;
1188 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
1189 __u32 pb_service_time; /* for rep, actual service time, also used for
1190 net_latency of req */
1193 /* VBR: pre-versions */
1194 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
1195 /* padding for future needs */
1196 __u64 pb_padding[4];
1199 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
1201 /* message body offset for lustre_msg_v2 */
1202 /* ptlrpc body offset in all request/reply messages */
1203 #define MSG_PTLRPC_BODY_OFF 0
1205 /* normal request/reply message record offset */
1206 #define REQ_REC_OFF 1
1207 #define REPLY_REC_OFF 1
1209 /* ldlm request message body offset */
1210 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
1211 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
1213 /* ldlm intent lock message body offset */
1214 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
1215 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
1217 /* ldlm reply message body offset */
1218 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
1219 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
1221 /** only use in req->rq_{req,rep}_swab_mask */
1222 #define MSG_PTLRPC_HEADER_OFF 31
1224 /* Flags that are operation-specific go in the top 16 bits. */
1225 #define MSG_OP_FLAG_MASK 0xffff0000
1226 #define MSG_OP_FLAG_SHIFT 16
1228 /* Flags that apply to all requests are in the bottom 16 bits */
1229 #define MSG_GEN_FLAG_MASK 0x0000ffff
1230 #define MSG_LAST_REPLAY 0x0001
1231 #define MSG_RESENT 0x0002
1232 #define MSG_REPLAY 0x0004
1233 /* #define MSG_AT_SUPPORT 0x0008
1234 * This was used in early prototypes of adaptive timeouts, and while there
1235 * shouldn't be any users of that code there also isn't a need for using this
1236 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
1237 #define MSG_DELAY_REPLAY 0x0010
1238 #define MSG_VERSION_REPLAY 0x0020
1239 #define MSG_REQ_REPLAY_DONE 0x0040
1240 #define MSG_LOCK_REPLAY_DONE 0x0080
1243 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
1246 #define MSG_CONNECT_RECOVERING 0x00000001
1247 #define MSG_CONNECT_RECONNECT 0x00000002
1248 #define MSG_CONNECT_REPLAYABLE 0x00000004
1249 //#define MSG_CONNECT_PEER 0x8
1250 #define MSG_CONNECT_LIBCLIENT 0x00000010
1251 #define MSG_CONNECT_INITIAL 0x00000020
1252 #define MSG_CONNECT_ASYNC 0x00000040
1253 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
1254 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
1257 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
1258 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
1259 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
1260 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
1261 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
1262 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
1263 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
1264 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
1265 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
1266 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create obj on write*/
1267 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
1268 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
1269 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks*/
1270 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated.
1271 *We do not support JOIN FILE
1272 *anymore, reserve this flags
1273 *just for preventing such bit
1275 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
1276 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
1277 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /*Remote client */
1278 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /*Remote client by force */
1279 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
1280 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
1281 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
1282 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
1283 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
1284 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
1285 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
1286 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
1287 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
1288 #define OBD_CONNECT_REAL 0x8000000ULL /*real connection */
1289 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
1290 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
1291 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
1292 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
1293 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
1294 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
1295 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
1296 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
1297 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
1298 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
1299 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
1301 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
1302 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
1303 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
1304 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
1305 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
1306 * RPC error properly */
1307 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
1308 * finer space reservation */
1309 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
1310 * policy and 2.x server */
1311 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
1312 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
1313 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
1314 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
1315 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
1317 * Please DO NOT add flag values here before first ensuring that this same
1318 * flag value is not in use on some other branch. Please clear any such
1319 * changes with senior engineers before starting to use a new flag. Then,
1320 * submit a small patch against EVERY branch that ONLY adds the new flag,
1321 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
1322 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
1323 * can be approved and landed easily to reserve the flag for future use. */
1325 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
1326 * connection. It is a temporary bug fix for Imperative Recovery interop
1327 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
1328 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
1329 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
1331 #define OCD_HAS_FLAG(ocd, flg) \
1332 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
1335 #ifdef HAVE_LRU_RESIZE_SUPPORT
1336 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
1338 #define LRU_RESIZE_CONNECT_FLAG 0
1341 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
1342 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
1343 OBD_CONNECT_IBITS | \
1344 OBD_CONNECT_NODEVOH | OBD_CONNECT_ATTRFID | \
1345 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1346 OBD_CONNECT_RMT_CLIENT | \
1347 OBD_CONNECT_RMT_CLIENT_FORCE | \
1348 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_MDS_CAPA | \
1349 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_MDS_MDS | \
1350 OBD_CONNECT_FID | LRU_RESIZE_CONNECT_FLAG | \
1351 OBD_CONNECT_VBR | OBD_CONNECT_LOV_V3 | \
1352 OBD_CONNECT_SOM | OBD_CONNECT_FULL20 | \
1353 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
1354 OBD_CONNECT_EINPROGRESS | \
1355 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
1356 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
1357 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE)
1358 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
1359 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
1360 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
1361 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_OSS_CAPA | \
1362 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
1363 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_CKSUM | \
1364 OBD_CONNECT_RMT_CLIENT | \
1365 OBD_CONNECT_RMT_CLIENT_FORCE | OBD_CONNECT_VBR | \
1366 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
1367 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 | \
1368 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
1369 OBD_CONNECT_MAX_EASIZE | \
1370 OBD_CONNECT_EINPROGRESS | \
1371 OBD_CONNECT_JOBSTATS | \
1372 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
1373 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
1374 OBD_CONNECT_PINGLESS)
1375 #define ECHO_CONNECT_SUPPORTED (0)
1376 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
1377 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
1378 OBD_CONNECT_MNE_SWAB | OBD_CONNECT_PINGLESS)
1380 /* Features required for this version of the client to work with server */
1381 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_IBITS | OBD_CONNECT_FID | \
1384 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
1385 ((patch)<<8) + (fix))
1386 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
1387 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
1388 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
1389 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
1391 /* This structure is used for both request and reply.
1393 * If we eventually have separate connect data for different types, which we
1394 * almost certainly will, then perhaps we stick a union in here. */
1395 struct obd_connect_data_v1 {
1396 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
1397 __u32 ocd_version; /* lustre release version number */
1398 __u32 ocd_grant; /* initial cache grant amount (bytes) */
1399 __u32 ocd_index; /* LOV index to connect to */
1400 __u32 ocd_brw_size; /* Maximum BRW size in bytes, must be 2^n */
1401 __u64 ocd_ibits_known; /* inode bits this client understands */
1402 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
1403 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
1404 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
1405 __u32 ocd_unused; /* also fix lustre_swab_connect */
1406 __u64 ocd_transno; /* first transno from client to be replayed */
1407 __u32 ocd_group; /* MDS group on OST */
1408 __u32 ocd_cksum_types; /* supported checksum algorithms */
1409 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1410 __u32 ocd_instance; /* also fix lustre_swab_connect */
1411 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1414 struct obd_connect_data {
1415 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
1416 __u32 ocd_version; /* lustre release version number */
1417 __u32 ocd_grant; /* initial cache grant amount (bytes) */
1418 __u32 ocd_index; /* LOV index to connect to */
1419 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
1420 __u64 ocd_ibits_known; /* inode bits this client understands */
1421 __u8 ocd_blocksize; /* log2 of the backend filesystem blocksize */
1422 __u8 ocd_inodespace; /* log2 of the per-inode space consumption */
1423 __u16 ocd_grant_extent; /* per-extent grant overhead, in 1K blocks */
1424 __u32 ocd_unused; /* also fix lustre_swab_connect */
1425 __u64 ocd_transno; /* first transno from client to be replayed */
1426 __u32 ocd_group; /* MDS group on OST */
1427 __u32 ocd_cksum_types; /* supported checksum algorithms */
1428 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1429 __u32 ocd_instance; /* instance # of this target */
1430 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1431 /* Fields after ocd_maxbytes are only accessible by the receiver
1432 * if the corresponding flag in ocd_connect_flags is set. Accessing
1433 * any field after ocd_maxbytes on the receiver without a valid flag
1434 * may result in out-of-bound memory access and kernel oops. */
1435 __u64 padding1; /* added 2.1.0. also fix lustre_swab_connect */
1436 __u64 padding2; /* added 2.1.0. also fix lustre_swab_connect */
1437 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
1438 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
1439 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
1440 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
1441 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
1442 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
1443 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
1444 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
1445 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
1446 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
1447 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
1448 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
1449 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1452 * Please DO NOT use any fields here before first ensuring that this same
1453 * field is not in use on some other branch. Please clear any such changes
1454 * with senior engineers before starting to use a new field. Then, submit
1455 * a small patch against EVERY branch that ONLY adds the new field along with
1456 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
1457 * reserve the flag for future use. */
1460 extern void lustre_swab_connect(struct obd_connect_data *ocd);
1463 * Supported checksum algorithms. Up to 32 checksum types are supported.
1464 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1465 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
1466 * algorithm and also the OBD_FL_CKSUM* flags.
1469 OBD_CKSUM_CRC32 = 0x00000001,
1470 OBD_CKSUM_ADLER = 0x00000002,
1471 OBD_CKSUM_CRC32C= 0x00000004,
1475 * OST requests: OBDO & OBD request records
1480 OST_REPLY = 0, /* reply ? */
1496 OST_QUOTACHECK = 18,
1498 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1501 #define OST_FIRST_OPC OST_REPLY
1504 OBD_FL_INLINEDATA = 0x00000001,
1505 OBD_FL_OBDMDEXISTS = 0x00000002,
1506 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1507 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1508 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1509 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1510 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1511 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1512 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1513 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1514 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1515 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1516 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1517 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1518 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
1519 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1520 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1521 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1522 * XXX: obsoleted - reserved for old
1523 * clients prior than 2.2 */
1524 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1525 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1527 /* Note that while these checksum values are currently separate bits,
1528 * in 2.x we can actually allow all values from 1-31 if we wanted. */
1529 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1530 OBD_FL_CKSUM_CRC32C,
1532 /* mask for local-only flag, which won't be sent over network */
1533 OBD_FL_LOCAL_MASK = 0xF0000000,
1536 #define LOV_MAGIC_V1 0x0BD10BD0
1537 #define LOV_MAGIC LOV_MAGIC_V1
1538 #define LOV_MAGIC_JOIN_V1 0x0BD20BD0
1539 #define LOV_MAGIC_V3 0x0BD30BD0
1542 * magic for fully defined striping
1543 * the idea is that we should have different magics for striping "hints"
1544 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1545 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1546 * we can't just change it w/o long way preparation, but we still need a
1547 * mechanism to allow LOD to differentiate hint versus ready striping.
1548 * so, at the moment we do a trick: MDT knows what to expect from request
1549 * depending on the case (replay uses ready striping, non-replay req uses
1550 * hints), so MDT replaces magic with appropriate one and now LOD can
1551 * easily understand what's inside -bzzz
1553 #define LOV_MAGIC_V1_DEF 0x0CD10BD0
1554 #define LOV_MAGIC_V3_DEF 0x0CD30BD0
1556 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
1557 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
1558 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
1559 #define LOV_PATTERN_CMOBD 0x200
1561 #define LOV_PATTERN_F_MASK 0xffff0000
1562 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
1564 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1565 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1567 #define lov_ost_data lov_ost_data_v1
1568 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1569 struct ost_id l_ost_oi; /* OST object ID */
1570 __u32 l_ost_gen; /* generation of this l_ost_idx */
1571 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1574 #define lov_mds_md lov_mds_md_v1
1575 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1576 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1577 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1578 struct ost_id lmm_oi; /* LOV object ID */
1579 __u32 lmm_stripe_size; /* size of stripe in bytes */
1580 /* lmm_stripe_count used to be __u32 */
1581 __u16 lmm_stripe_count; /* num stripes in use for this object */
1582 __u16 lmm_layout_gen; /* layout generation number */
1583 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1587 * Sigh, because pre-2.4 uses
1588 * struct lov_mds_md_v1 {
1590 * __u64 lmm_object_id;
1591 * __u64 lmm_object_seq;
1594 * to identify the LOV(MDT) object, and lmm_object_seq will
1595 * be normal_fid, which make it hard to combine these conversion
1596 * to ostid_to FID. so we will do lmm_oi/fid conversion separately
1598 * We can tell the lmm_oi by this way,
1599 * 1.8: lmm_object_id = {inode}, lmm_object_gr = 0
1600 * 2.1: lmm_object_id = {oid < 128k}, lmm_object_seq = FID_SEQ_NORMAL
1601 * 2.4: lmm_oi.f_seq = FID_SEQ_NORMAL, lmm_oi.f_oid = {oid < 128k},
1604 * But currently lmm_oi/lsm_oi does not have any "real" usages,
1605 * except for printing some information, and the user can always
1606 * get the real FID from LMA, besides this multiple case check might
1607 * make swab more complicate. So we will keep using id/seq for lmm_oi.
1610 static inline void fid_to_lmm_oi(const struct lu_fid *fid,
1613 oi->oi.oi_id = fid_oid(fid);
1614 oi->oi.oi_seq = fid_seq(fid);
1617 static inline void lmm_oi_set_seq(struct ost_id *oi, __u64 seq)
1619 oi->oi.oi_seq = seq;
1622 static inline __u64 lmm_oi_id(struct ost_id *oi)
1624 return oi->oi.oi_id;
1627 static inline __u64 lmm_oi_seq(struct ost_id *oi)
1629 return oi->oi.oi_seq;
1632 static inline void lmm_oi_le_to_cpu(struct ost_id *dst_oi,
1633 struct ost_id *src_oi)
1635 dst_oi->oi.oi_id = le64_to_cpu(src_oi->oi.oi_id);
1636 dst_oi->oi.oi_seq = le64_to_cpu(src_oi->oi.oi_seq);
1639 static inline void lmm_oi_cpu_to_le(struct ost_id *dst_oi,
1640 struct ost_id *src_oi)
1642 dst_oi->oi.oi_id = cpu_to_le64(src_oi->oi.oi_id);
1643 dst_oi->oi.oi_seq = cpu_to_le64(src_oi->oi.oi_seq);
1646 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
1648 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1649 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1651 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1652 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1653 #define XATTR_USER_PREFIX "user."
1654 #define XATTR_TRUSTED_PREFIX "trusted."
1655 #define XATTR_SECURITY_PREFIX "security."
1656 #define XATTR_LUSTRE_PREFIX "lustre."
1658 #define XATTR_NAME_LOV "trusted.lov"
1659 #define XATTR_NAME_LMA "trusted.lma"
1660 #define XATTR_NAME_LMV "trusted.lmv"
1661 #define XATTR_NAME_LINK "trusted.link"
1662 #define XATTR_NAME_FID "trusted.fid"
1663 #define XATTR_NAME_VERSION "trusted.version"
1664 #define XATTR_NAME_SOM "trusted.som"
1665 #define XATTR_NAME_HSM "trusted.hsm"
1666 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_namespace"
1668 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1669 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1670 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1671 struct ost_id lmm_oi; /* LOV object ID */
1672 __u32 lmm_stripe_size; /* size of stripe in bytes */
1673 /* lmm_stripe_count used to be __u32 */
1674 __u16 lmm_stripe_count; /* num stripes in use for this object */
1675 __u16 lmm_layout_gen; /* layout generation number */
1676 char lmm_pool_name[LOV_MAXPOOLNAME]; /* must be 32bit aligned */
1677 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1680 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1682 if (lmm_magic == LOV_MAGIC_V3)
1683 return sizeof(struct lov_mds_md_v3) +
1684 stripes * sizeof(struct lov_ost_data_v1);
1686 return sizeof(struct lov_mds_md_v1) +
1687 stripes * sizeof(struct lov_ost_data_v1);
1691 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1692 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1693 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1694 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1695 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1696 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1697 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1698 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1699 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1700 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1701 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1702 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1703 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1704 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1705 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
1706 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1707 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1708 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1709 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1710 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1711 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1712 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
1713 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
1714 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1715 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1716 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1717 /* ->mds if epoch opens or closes */
1718 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1719 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1720 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1721 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1722 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1724 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1725 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1726 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1727 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1729 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1730 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1731 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1732 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1733 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
1734 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1735 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1736 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1737 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1738 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1739 * under lock; for xattr
1740 * requests means the
1741 * client holds the lock */
1742 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1744 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
1745 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
1746 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
1747 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
1749 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1750 #define OBD_MD_FLRELEASED (0x0020000000000000ULL) /* file released */
1752 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1753 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1754 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1755 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1756 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
1758 #define OBD_MD_FLXATTRLOCKED OBD_MD_FLGETATTRLOCK
1759 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1761 /* don't forget obdo_fid which is way down at the bottom so it can
1762 * come after the definition of llog_cookie */
1766 HSS_CLEARMASK = 0x02,
1767 HSS_ARCHIVE_ID = 0x04,
1770 struct hsm_state_set {
1772 __u32 hss_archive_id;
1774 __u64 hss_clearmask;
1777 extern void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
1778 extern void lustre_swab_hsm_state_set(struct hsm_state_set *hss);
1780 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
1782 /* ost_body.data values for OST_BRW */
1784 #define OBD_BRW_READ 0x01
1785 #define OBD_BRW_WRITE 0x02
1786 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1787 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1788 * transfer and is not accounted in
1790 #define OBD_BRW_CHECK 0x10
1791 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1792 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1793 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1794 #define OBD_BRW_NOQUOTA 0x100
1795 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1796 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1797 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1798 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1799 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1801 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
1803 #define OST_MIN_PRECREATE 32
1804 #define OST_MAX_PRECREATE 20000
1807 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1808 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1809 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1810 * high 16 bits in 2.4 and later */
1811 __u32 ioo_bufcnt; /* number of niobufs for this object */
1814 #define IOOBJ_MAX_BRW_BITS 16
1815 #define IOOBJ_TYPE_MASK ((1U << IOOBJ_MAX_BRW_BITS) - 1)
1816 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1817 #define ioobj_max_brw_set(ioo, num) \
1818 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1820 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
1822 /* multiple of 8 bytes => can array */
1823 struct niobuf_remote {
1829 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
1831 /* lock value block communicated between the filter and llite */
1833 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1834 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1835 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1836 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1837 #define OST_LVB_IS_ERR(blocks) \
1838 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1839 #define OST_LVB_SET_ERR(blocks, rc) \
1840 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1841 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1851 extern void lustre_swab_ost_lvb_v1(struct ost_lvb_v1 *lvb);
1865 extern void lustre_swab_ost_lvb(struct ost_lvb *lvb);
1868 * lquota data structures
1871 #ifndef QUOTABLOCK_BITS
1872 #define QUOTABLOCK_BITS 10
1875 #ifndef QUOTABLOCK_SIZE
1876 #define QUOTABLOCK_SIZE (1 << QUOTABLOCK_BITS)
1880 #define toqb(x) (((x) + QUOTABLOCK_SIZE - 1) >> QUOTABLOCK_BITS)
1883 /* The lquota_id structure is an union of all the possible identifier types that
1884 * can be used with quota, this includes:
1887 * - a FID which can be used for per-directory quota in the future */
1889 struct lu_fid qid_fid; /* FID for per-directory quota */
1890 __u64 qid_uid; /* user identifier */
1891 __u64 qid_gid; /* group identifier */
1894 /* quotactl management */
1895 struct obd_quotactl {
1897 __u32 qc_type; /* see Q_* flag below */
1900 struct obd_dqinfo qc_dqinfo;
1901 struct obd_dqblk qc_dqblk;
1904 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1906 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1907 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1908 #define Q_GETOINFO 0x800102 /* get obd quota info */
1909 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1910 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1912 #define Q_COPY(out, in, member) (out)->member = (in)->member
1914 #define QCTL_COPY(out, in) \
1916 Q_COPY(out, in, qc_cmd); \
1917 Q_COPY(out, in, qc_type); \
1918 Q_COPY(out, in, qc_id); \
1919 Q_COPY(out, in, qc_stat); \
1920 Q_COPY(out, in, qc_dqinfo); \
1921 Q_COPY(out, in, qc_dqblk); \
1924 /* Body of quota request used for quota acquire/release RPCs between quota
1925 * master (aka QMT) and slaves (ak QSD). */
1927 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1928 * and type (data or metadata) as well as
1929 * the quota type (user or group). */
1930 union lquota_id qb_id; /* uid or gid or directory FID */
1931 __u32 qb_flags; /* see below */
1933 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1934 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1935 __u64 qb_slv_ver; /* slave index file version */
1936 struct lustre_handle qb_lockh; /* per-ID lock handle */
1937 struct lustre_handle qb_glb_lockh; /* global lock handle */
1938 __u64 qb_padding1[4];
1941 /* When the quota_body is used in the reply of quota global intent
1942 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1943 #define qb_slv_fid qb_fid
1944 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1946 #define qb_qunit qb_usage
1948 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1949 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1950 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1951 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1953 extern void lustre_swab_quota_body(struct quota_body *b);
1955 /* Quota types currently supported */
1957 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1958 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1962 /* There are 2 different resource types on which a quota limit can be enforced:
1963 * - inodes on the MDTs
1964 * - blocks on the OSTs */
1966 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1967 LQUOTA_RES_DT = 0x02,
1969 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1971 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1974 * Space accounting support
1975 * Format of an accounting record, providing disk usage information for a given
1978 struct lquota_acct_rec { /* 16 bytes */
1979 __u64 bspace; /* current space in use */
1980 __u64 ispace; /* current # inodes in use */
1984 * Global quota index support
1985 * Format of a global record, providing global quota settings for a given quota
1988 struct lquota_glb_rec { /* 32 bytes */
1989 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1990 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1991 __u64 qbr_time; /* grace time, in seconds */
1992 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1997 * Slave index support
1998 * Format of a slave record, recording how much space is granted to a given
2001 struct lquota_slv_rec { /* 8 bytes */
2002 __u64 qsr_granted; /* space granted to the slave for the key=ID,
2003 * in #inodes or kbytes */
2006 /* Data structures associated with the quota locks */
2008 /* Glimpse descriptor used for the index & per-ID quota locks */
2009 struct ldlm_gl_lquota_desc {
2010 union lquota_id gl_id; /* quota ID subject to the glimpse */
2011 __u64 gl_flags; /* see LQUOTA_FL* below */
2012 __u64 gl_ver; /* new index version */
2013 __u64 gl_hardlimit; /* new hardlimit or qunit value */
2014 __u64 gl_softlimit; /* new softlimit */
2018 #define gl_qunit gl_hardlimit /* current qunit value used when
2019 * glimpsing per-ID quota locks */
2021 /* quota glimpse flags */
2022 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
2024 /* LVB used with quota (global and per-ID) locks */
2026 __u64 lvb_flags; /* see LQUOTA_FL* above */
2027 __u64 lvb_id_may_rel; /* space that might be released later */
2028 __u64 lvb_id_rel; /* space released by the slave for this ID */
2029 __u64 lvb_id_qunit; /* current qunit value */
2033 extern void lustre_swab_lquota_lvb(struct lquota_lvb *lvb);
2035 /* LVB used with global quota lock */
2036 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
2044 #define QUOTA_FIRST_OPC QUOTA_DQACQ
2053 MDS_GETATTR_NAME = 34,
2058 MDS_DISCONNECT = 39,
2064 MDS_DONE_WRITING = 45,
2066 MDS_QUOTACHECK = 47,
2069 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
2073 MDS_HSM_STATE_GET = 54,
2074 MDS_HSM_STATE_SET = 55,
2075 MDS_HSM_ACTION = 56,
2076 MDS_HSM_PROGRESS = 57,
2077 MDS_HSM_REQUEST = 58,
2078 MDS_HSM_CT_REGISTER = 59,
2079 MDS_HSM_CT_UNREGISTER = 60,
2080 MDS_SWAP_LAYOUTS = 61,
2084 #define MDS_FIRST_OPC MDS_GETATTR
2087 /* opcodes for object update */
2093 #define UPDATE_FIRST_OPC UPDATE_OBJ
2110 } mds_reint_t, mdt_reint_t;
2112 extern void lustre_swab_generic_32s (__u32 *val);
2114 /* the disposition of the intent outlines what was executed */
2115 #define DISP_IT_EXECD 0x00000001
2116 #define DISP_LOOKUP_EXECD 0x00000002
2117 #define DISP_LOOKUP_NEG 0x00000004
2118 #define DISP_LOOKUP_POS 0x00000008
2119 #define DISP_OPEN_CREATE 0x00000010
2120 #define DISP_OPEN_OPEN 0x00000020
2121 #define DISP_ENQ_COMPLETE 0x00400000
2122 #define DISP_ENQ_OPEN_REF 0x00800000
2123 #define DISP_ENQ_CREATE_REF 0x01000000
2124 #define DISP_OPEN_LOCK 0x02000000
2125 #define DISP_OPEN_LEASE 0x04000000
2127 /* INODE LOCK PARTS */
2128 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
2129 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
2130 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
2131 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
2132 #define MDS_INODELOCK_PERM 0x000010 /* for permission */
2133 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
2135 #define MDS_INODELOCK_MAXSHIFT 5
2136 /* This FULL lock is useful to take on unlink sort of operations */
2137 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
2139 extern void lustre_swab_ll_fid (struct ll_fid *fid);
2141 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
2142 * but was moved into name[1] along with the OID to avoid consuming the
2143 * name[2,3] fields that need to be used for the quota id (also a FID). */
2145 LUSTRE_RES_ID_SEQ_OFF = 0,
2146 LUSTRE_RES_ID_VER_OID_OFF = 1,
2147 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
2148 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
2149 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
2150 LUSTRE_RES_ID_HSH_OFF = 3
2153 #define MDS_STATUS_CONN 1
2154 #define MDS_STATUS_LOV 2
2156 /* mdt_thread_info.mti_flags. */
2158 /* The flag indicates Size-on-MDS attributes are changed. */
2159 MF_SOM_CHANGE = (1 << 0),
2160 /* Flags indicates an epoch opens or closes. */
2161 MF_EPOCH_OPEN = (1 << 1),
2162 MF_EPOCH_CLOSE = (1 << 2),
2163 MF_MDC_CANCEL_FID1 = (1 << 3),
2164 MF_MDC_CANCEL_FID2 = (1 << 4),
2165 MF_MDC_CANCEL_FID3 = (1 << 5),
2166 MF_MDC_CANCEL_FID4 = (1 << 6),
2167 /* There is a pending attribute update. */
2168 MF_SOM_AU = (1 << 7),
2169 /* Cancel OST locks while getattr OST attributes. */
2170 MF_GETATTR_LOCK = (1 << 8),
2171 MF_GET_MDT_IDX = (1 << 9),
2174 #define MF_SOM_LOCAL_FLAGS (MF_SOM_CHANGE | MF_EPOCH_OPEN | MF_EPOCH_CLOSE)
2176 #define LUSTRE_BFLAG_UNCOMMITTED_WRITES 0x1
2178 /* these should be identical to their EXT4_*_FL counterparts, they are
2179 * redefined here only to avoid dragging in fs/ext4/ext4.h */
2180 #define LUSTRE_SYNC_FL 0x00000008 /* Synchronous updates */
2181 #define LUSTRE_IMMUTABLE_FL 0x00000010 /* Immutable file */
2182 #define LUSTRE_APPEND_FL 0x00000020 /* writes to file may only append */
2183 #define LUSTRE_NOATIME_FL 0x00000080 /* do not update atime */
2184 #define LUSTRE_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
2187 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
2188 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
2189 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
2190 * the S_* flags are kernel-internal values that change between kernel
2191 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
2192 * See b=16526 for a full history. */
2193 static inline int ll_ext_to_inode_flags(int flags)
2195 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
2196 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
2197 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
2198 #if defined(S_DIRSYNC)
2199 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
2201 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0));
2204 static inline int ll_inode_to_ext_flags(int iflags)
2206 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
2207 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
2208 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
2209 #if defined(S_DIRSYNC)
2210 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
2212 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0));
2216 /* 64 possible states */
2217 enum md_transient_state {
2218 MS_RESTORE = (1 << 0), /* restore is running */
2224 struct lustre_handle handle;
2226 __u64 size; /* Offset, in the case of MDS_READPAGE */
2230 __u64 blocks; /* XID, in the case of MDS_READPAGE */
2232 __u64 t_state; /* transient file state defined in
2233 * enum md_transient_state
2234 * was "ino" until 2.4.0 */
2241 __u32 flags; /* from vfs for pin/unpin, LUSTRE_BFLAG close */
2243 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
2244 __u32 unused2; /* was "generation" until 2.4.0 */
2249 __u32 max_cookiesize;
2250 __u32 uid_h; /* high 32-bits of uid, for FUID */
2251 __u32 gid_h; /* high 32-bits of gid, for FUID */
2252 __u32 padding_5; /* also fix lustre_swab_mdt_body */
2260 extern void lustre_swab_mdt_body (struct mdt_body *b);
2262 struct mdt_ioepoch {
2263 struct lustre_handle handle;
2269 extern void lustre_swab_mdt_ioepoch (struct mdt_ioepoch *b);
2271 /* permissions for md_perm.mp_perm */
2273 CFS_SETUID_PERM = 0x01,
2274 CFS_SETGID_PERM = 0x02,
2275 CFS_SETGRP_PERM = 0x04,
2276 CFS_RMTACL_PERM = 0x08,
2277 CFS_RMTOWN_PERM = 0x10
2280 /* inode access permission for remote user, the inode info are omitted,
2281 * for client knows them. */
2282 struct mdt_remote_perm {
2289 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
2293 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
2295 struct mdt_rec_setattr {
2305 __u32 sa_padding_1_h;
2306 struct lu_fid sa_fid;
2315 __u32 sa_attr_flags;
2317 __u32 sa_bias; /* some operation flags */
2323 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
2326 * Attribute flags used in mdt_rec_setattr::sa_valid.
2327 * The kernel's #defines for ATTR_* should not be used over the network
2328 * since the client and MDS may run different kernels (see bug 13828)
2329 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
2331 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
2332 #define MDS_ATTR_UID 0x2ULL /* = 2 */
2333 #define MDS_ATTR_GID 0x4ULL /* = 4 */
2334 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
2335 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
2336 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
2337 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
2338 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
2339 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
2340 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
2341 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
2342 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
2343 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
2344 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
2345 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
2346 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
2349 #define FMODE_READ 00000001
2350 #define FMODE_WRITE 00000002
2353 #define MDS_FMODE_CLOSED 00000000
2354 #define MDS_FMODE_EXEC 00000004
2355 /* IO Epoch is opened on a closed file. */
2356 #define MDS_FMODE_EPOCH 01000000
2357 /* IO Epoch is opened on a file truncate. */
2358 #define MDS_FMODE_TRUNC 02000000
2359 /* Size-on-MDS Attribute Update is pending. */
2360 #define MDS_FMODE_SOM 04000000
2362 #define MDS_OPEN_CREATED 00000010
2363 #define MDS_OPEN_CROSS 00000020
2365 #define MDS_OPEN_CREAT 00000100
2366 #define MDS_OPEN_EXCL 00000200
2367 #define MDS_OPEN_TRUNC 00001000
2368 #define MDS_OPEN_APPEND 00002000
2369 #define MDS_OPEN_SYNC 00010000
2370 #define MDS_OPEN_DIRECTORY 00200000
2372 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
2373 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
2374 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
2375 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
2376 * We do not support JOIN FILE
2377 * anymore, reserve this flags
2378 * just for preventing such bit
2381 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
2382 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
2383 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
2384 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
2385 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
2387 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
2389 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
2390 * delegation, succeed if it's not
2391 * being opened with conflict mode.
2393 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
2395 /* permission for create non-directory file */
2396 #define MAY_CREATE (1 << 7)
2397 /* permission for create directory file */
2398 #define MAY_LINK (1 << 8)
2399 /* permission for delete from the directory */
2400 #define MAY_UNLINK (1 << 9)
2401 /* source's permission for rename */
2402 #define MAY_RENAME_SRC (1 << 10)
2403 /* target's permission for rename */
2404 #define MAY_RENAME_TAR (1 << 11)
2405 /* part (parent's) VTX permission check */
2406 #define MAY_VTX_PART (1 << 12)
2407 /* full VTX permission check */
2408 #define MAY_VTX_FULL (1 << 13)
2409 /* lfs rgetfacl permission check */
2410 #define MAY_RGETFACL (1 << 14)
2413 MDS_CHECK_SPLIT = 1 << 0,
2414 MDS_CROSS_REF = 1 << 1,
2415 MDS_VTX_BYPASS = 1 << 2,
2416 MDS_PERM_BYPASS = 1 << 3,
2418 MDS_QUOTA_IGNORE = 1 << 5,
2419 MDS_CLOSE_CLEANUP = 1 << 6,
2420 MDS_KEEP_ORPHAN = 1 << 7,
2421 MDS_RECOV_OPEN = 1 << 8,
2422 MDS_DATA_MODIFIED = 1 << 9,
2423 MDS_CREATE_VOLATILE = 1 << 10,
2424 MDS_OWNEROVERRIDE = 1 << 11,
2425 MDS_HSM_RELEASE = 1 << 12,
2428 /* instance of mdt_reint_rec */
2429 struct mdt_rec_create {
2437 __u32 cr_suppgid1_h;
2439 __u32 cr_suppgid2_h;
2440 struct lu_fid cr_fid1;
2441 struct lu_fid cr_fid2;
2442 struct lustre_handle cr_old_handle; /* handle in case of open replay */
2446 __u64 cr_padding_1; /* rr_blocks */
2449 /* use of helpers set/get_mrc_cr_flags() is needed to access
2450 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2451 * extend cr_flags size without breaking 1.8 compat */
2452 __u32 cr_flags_l; /* for use with open, low 32 bits */
2453 __u32 cr_flags_h; /* for use with open, high 32 bits */
2454 __u32 cr_umask; /* umask for create */
2455 __u32 cr_padding_4; /* rr_padding_4 */
2458 static inline void set_mrc_cr_flags(struct mdt_rec_create *mrc, __u64 flags)
2460 mrc->cr_flags_l = (__u32)(flags & 0xFFFFFFFFUll);
2461 mrc->cr_flags_h = (__u32)(flags >> 32);
2464 static inline __u64 get_mrc_cr_flags(struct mdt_rec_create *mrc)
2466 return ((__u64)(mrc->cr_flags_l) | ((__u64)mrc->cr_flags_h << 32));
2469 /* instance of mdt_reint_rec */
2470 struct mdt_rec_link {
2478 __u32 lk_suppgid1_h;
2480 __u32 lk_suppgid2_h;
2481 struct lu_fid lk_fid1;
2482 struct lu_fid lk_fid2;
2484 __u64 lk_padding_1; /* rr_atime */
2485 __u64 lk_padding_2; /* rr_ctime */
2486 __u64 lk_padding_3; /* rr_size */
2487 __u64 lk_padding_4; /* rr_blocks */
2489 __u32 lk_padding_5; /* rr_mode */
2490 __u32 lk_padding_6; /* rr_flags */
2491 __u32 lk_padding_7; /* rr_padding_2 */
2492 __u32 lk_padding_8; /* rr_padding_3 */
2493 __u32 lk_padding_9; /* rr_padding_4 */
2496 /* instance of mdt_reint_rec */
2497 struct mdt_rec_unlink {
2505 __u32 ul_suppgid1_h;
2507 __u32 ul_suppgid2_h;
2508 struct lu_fid ul_fid1;
2509 struct lu_fid ul_fid2;
2511 __u64 ul_padding_2; /* rr_atime */
2512 __u64 ul_padding_3; /* rr_ctime */
2513 __u64 ul_padding_4; /* rr_size */
2514 __u64 ul_padding_5; /* rr_blocks */
2517 __u32 ul_padding_6; /* rr_flags */
2518 __u32 ul_padding_7; /* rr_padding_2 */
2519 __u32 ul_padding_8; /* rr_padding_3 */
2520 __u32 ul_padding_9; /* rr_padding_4 */
2523 /* instance of mdt_reint_rec */
2524 struct mdt_rec_rename {
2532 __u32 rn_suppgid1_h;
2534 __u32 rn_suppgid2_h;
2535 struct lu_fid rn_fid1;
2536 struct lu_fid rn_fid2;
2538 __u64 rn_padding_1; /* rr_atime */
2539 __u64 rn_padding_2; /* rr_ctime */
2540 __u64 rn_padding_3; /* rr_size */
2541 __u64 rn_padding_4; /* rr_blocks */
2542 __u32 rn_bias; /* some operation flags */
2543 __u32 rn_mode; /* cross-ref rename has mode */
2544 __u32 rn_padding_5; /* rr_flags */
2545 __u32 rn_padding_6; /* rr_padding_2 */
2546 __u32 rn_padding_7; /* rr_padding_3 */
2547 __u32 rn_padding_8; /* rr_padding_4 */
2550 /* instance of mdt_reint_rec */
2551 struct mdt_rec_setxattr {
2559 __u32 sx_suppgid1_h;
2561 __u32 sx_suppgid2_h;
2562 struct lu_fid sx_fid;
2563 __u64 sx_padding_1; /* These three are rr_fid2 */
2568 __u64 sx_padding_5; /* rr_ctime */
2569 __u64 sx_padding_6; /* rr_size */
2570 __u64 sx_padding_7; /* rr_blocks */
2573 __u32 sx_padding_8; /* rr_flags */
2574 __u32 sx_padding_9; /* rr_padding_2 */
2575 __u32 sx_padding_10; /* rr_padding_3 */
2576 __u32 sx_padding_11; /* rr_padding_4 */
2580 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2581 * Do NOT change the size of various members, otherwise the value
2582 * will be broken in lustre_swab_mdt_rec_reint().
2584 * If you add new members in other mdt_reint_xxx structres and need to use the
2585 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2587 struct mdt_rec_reint {
2595 __u32 rr_suppgid1_h;
2597 __u32 rr_suppgid2_h;
2598 struct lu_fid rr_fid1;
2599 struct lu_fid rr_fid2;
2610 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2613 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
2616 __u32 ld_tgt_count; /* how many MDS's */
2617 __u32 ld_active_tgt_count; /* how many active */
2618 __u32 ld_default_stripe_count; /* how many objects are used */
2619 __u32 ld_pattern; /* default MEA_MAGIC_* */
2620 __u64 ld_default_hash_size;
2621 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2622 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2623 __u32 ld_qos_maxage; /* in second */
2624 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2625 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2626 struct obd_uuid ld_uuid;
2629 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
2631 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
2632 struct lmv_stripe_md {
2637 char mea_pool_name[LOV_MAXPOOLNAME];
2638 struct lu_fid mea_ids[0];
2641 extern void lustre_swab_lmv_stripe_md(struct lmv_stripe_md *mea);
2643 /* lmv structures */
2644 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
2645 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
2646 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
2648 #define MAX_HASH_SIZE_32 0x7fffffffUL
2649 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
2650 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
2655 FLD_FIRST_OPC = FLD_QUERY
2661 SEQ_FIRST_OPC = SEQ_QUERY
2665 SEQ_ALLOC_SUPER = 0,
2670 * LOV data structures
2673 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2674 /* The size of the buffer the lov/mdc reserves for the
2675 * array of UUIDs returned by the MDS. With the current
2676 * protocol, this will limit the max number of OSTs per LOV */
2678 #define LOV_DESC_MAGIC 0xB0CCDE5C
2680 /* LOV settings descriptor (should only contain static info) */
2682 __u32 ld_tgt_count; /* how many OBD's */
2683 __u32 ld_active_tgt_count; /* how many active */
2684 __u32 ld_default_stripe_count; /* how many objects are used */
2685 __u32 ld_pattern; /* default PATTERN_RAID0 */
2686 __u64 ld_default_stripe_size; /* in bytes */
2687 __u64 ld_default_stripe_offset; /* in bytes */
2688 __u32 ld_padding_0; /* unused */
2689 __u32 ld_qos_maxage; /* in second */
2690 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2691 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2692 struct obd_uuid ld_uuid;
2695 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2697 extern void lustre_swab_lov_desc (struct lov_desc *ld);
2702 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2707 LDLM_BL_CALLBACK = 104,
2708 LDLM_CP_CALLBACK = 105,
2709 LDLM_GL_CALLBACK = 106,
2710 LDLM_SET_INFO = 107,
2713 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2715 #define RES_NAME_SIZE 4
2716 struct ldlm_res_id {
2717 __u64 name[RES_NAME_SIZE];
2720 #define DLDLMRES "["LPX64":"LPX64":"LPX64"]."LPX64i
2721 #define PLDLMRES(res) (res)->lr_name.name[0], (res)->lr_name.name[1], \
2722 (res)->lr_name.name[2], (res)->lr_name.name[3]
2724 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
2726 static inline int ldlm_res_eq(const struct ldlm_res_id *res0,
2727 const struct ldlm_res_id *res1)
2729 return !memcmp(res0, res1, sizeof(*res0));
2746 #define LCK_MODE_NUM 8
2756 #define LDLM_MIN_TYPE LDLM_PLAIN
2758 struct ldlm_extent {
2764 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
2765 struct ldlm_extent *ex2)
2767 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
2770 /* check if @ex1 contains @ex2 */
2771 static inline int ldlm_extent_contain(struct ldlm_extent *ex1,
2772 struct ldlm_extent *ex2)
2774 return (ex1->start <= ex2->start) && (ex1->end >= ex2->end);
2777 struct ldlm_inodebits {
2781 struct ldlm_flock_wire {
2789 /* it's important that the fields of the ldlm_extent structure match
2790 * the first fields of the ldlm_flock structure because there is only
2791 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2792 * this ever changes we will need to swab the union differently based
2793 * on the resource type. */
2796 struct ldlm_extent l_extent;
2797 struct ldlm_flock_wire l_flock;
2798 struct ldlm_inodebits l_inodebits;
2799 } ldlm_wire_policy_data_t;
2801 extern void lustre_swab_ldlm_policy_data (ldlm_wire_policy_data_t *d);
2803 union ldlm_gl_desc {
2804 struct ldlm_gl_lquota_desc lquota_desc;
2807 extern void lustre_swab_gl_desc(union ldlm_gl_desc *);
2809 struct ldlm_intent {
2813 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
2815 struct ldlm_resource_desc {
2816 ldlm_type_t lr_type;
2817 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
2818 struct ldlm_res_id lr_name;
2821 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
2823 struct ldlm_lock_desc {
2824 struct ldlm_resource_desc l_resource;
2825 ldlm_mode_t l_req_mode;
2826 ldlm_mode_t l_granted_mode;
2827 ldlm_wire_policy_data_t l_policy_data;
2830 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
2832 #define LDLM_LOCKREQ_HANDLES 2
2833 #define LDLM_ENQUEUE_CANCEL_OFF 1
2835 struct ldlm_request {
2838 struct ldlm_lock_desc lock_desc;
2839 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2842 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
2844 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
2845 * Otherwise, 2 are available. */
2846 #define ldlm_request_bufsize(count,type) \
2848 int _avail = LDLM_LOCKREQ_HANDLES; \
2849 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
2850 sizeof(struct ldlm_request) + \
2851 (count > _avail ? count - _avail : 0) * \
2852 sizeof(struct lustre_handle); \
2857 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2858 struct ldlm_lock_desc lock_desc;
2859 struct lustre_handle lock_handle;
2860 __u64 lock_policy_res1;
2861 __u64 lock_policy_res2;
2864 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
2866 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2867 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2870 * Opcodes for mountconf (mgs and mgc)
2875 MGS_EXCEPTION, /* node died, etc. */
2876 MGS_TARGET_REG, /* whenever target starts up */
2882 #define MGS_FIRST_OPC MGS_CONNECT
2884 #define MGS_PARAM_MAXLEN 1024
2885 #define KEY_SET_INFO "set_info"
2887 struct mgs_send_param {
2888 char mgs_param[MGS_PARAM_MAXLEN];
2891 /* We pass this info to the MGS so it can write config logs */
2892 #define MTI_NAME_MAXLEN 64
2893 #define MTI_PARAM_MAXLEN 4096
2894 #define MTI_NIDS_MAX 32
2895 struct mgs_target_info {
2896 __u32 mti_lustre_ver;
2897 __u32 mti_stripe_index;
2898 __u32 mti_config_ver;
2900 __u32 mti_nid_count;
2901 __u32 mti_instance; /* Running instance of target */
2902 char mti_fsname[MTI_NAME_MAXLEN];
2903 char mti_svname[MTI_NAME_MAXLEN];
2904 char mti_uuid[sizeof(struct obd_uuid)];
2905 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
2906 char mti_params[MTI_PARAM_MAXLEN];
2908 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
2910 struct mgs_nidtbl_entry {
2911 __u64 mne_version; /* table version of this entry */
2912 __u32 mne_instance; /* target instance # */
2913 __u32 mne_index; /* target index */
2914 __u32 mne_length; /* length of this entry - by bytes */
2915 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2916 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2917 __u8 mne_nid_size; /* size of each NID, by bytes */
2918 __u8 mne_nid_count; /* # of NIDs in buffer */
2920 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2923 extern void lustre_swab_mgs_nidtbl_entry(struct mgs_nidtbl_entry *oinfo);
2925 struct mgs_config_body {
2926 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2927 __u64 mcb_offset; /* next index of config log to request */
2928 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2930 __u8 mcb_bits; /* bits unit size of config log */
2931 __u32 mcb_units; /* # of units for bulk transfer */
2933 extern void lustre_swab_mgs_config_body(struct mgs_config_body *body);
2935 struct mgs_config_res {
2936 __u64 mcr_offset; /* index of last config log */
2937 __u64 mcr_size; /* size of the log */
2939 extern void lustre_swab_mgs_config_res(struct mgs_config_res *body);
2941 /* Config marker flags (in config log) */
2942 #define CM_START 0x01
2944 #define CM_SKIP 0x04
2945 #define CM_UPGRADE146 0x08
2946 #define CM_EXCLUDE 0x10
2947 #define CM_START_SKIP (CM_START | CM_SKIP)
2950 __u32 cm_step; /* aka config version */
2952 __u32 cm_vers; /* lustre release version number */
2953 __u32 cm_padding; /* 64 bit align */
2954 obd_time cm_createtime; /*when this record was first created */
2955 obd_time cm_canceltime; /*when this record is no longer valid*/
2956 char cm_tgtname[MTI_NAME_MAXLEN];
2957 char cm_comment[MTI_NAME_MAXLEN];
2960 extern void lustre_swab_cfg_marker(struct cfg_marker *marker,
2961 int swab, int size);
2964 * Opcodes for multiple servers.
2974 #define OBD_FIRST_OPC OBD_PING
2976 /* catalog of log objects */
2978 /** Identifier for a single log object */
2980 struct ost_id lgl_oi;
2982 } __attribute__((packed));
2984 /** Records written to the CATALOGS list */
2985 #define CATLIST "CATALOGS"
2987 struct llog_logid lci_logid;
2991 } __attribute__((packed));
2993 /* Log data record types - there is no specific reason that these need to
2994 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2996 #define LLOG_OP_MAGIC 0x10600000
2997 #define LLOG_OP_MASK 0xfff00000
3000 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
3001 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
3002 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
3003 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
3004 REINT_UNLINK, /* obsolete after 2.5.0 */
3005 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
3007 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
3008 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
3010 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
3011 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
3012 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
3013 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
3014 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
3015 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
3016 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
3017 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
3018 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
3021 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
3022 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
3024 /** Log record header - stored in little endian order.
3025 * Each record must start with this struct, end with a llog_rec_tail,
3026 * and be a multiple of 256 bits in size.
3028 struct llog_rec_hdr {
3035 struct llog_rec_tail {
3040 /* Where data follow just after header */
3041 #define REC_DATA(ptr) \
3042 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
3044 #define REC_DATA_LEN(rec) \
3045 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
3046 sizeof(struct llog_rec_tail))
3048 struct llog_logid_rec {
3049 struct llog_rec_hdr lid_hdr;
3050 struct llog_logid lid_id;
3054 struct llog_rec_tail lid_tail;
3055 } __attribute__((packed));
3057 struct llog_unlink_rec {
3058 struct llog_rec_hdr lur_hdr;
3061 obd_count lur_count;
3062 struct llog_rec_tail lur_tail;
3063 } __attribute__((packed));
3065 struct llog_unlink64_rec {
3066 struct llog_rec_hdr lur_hdr;
3067 struct lu_fid lur_fid;
3068 obd_count lur_count; /* to destroy the lost precreated */
3072 struct llog_rec_tail lur_tail;
3073 } __attribute__((packed));
3075 struct llog_setattr64_rec {
3076 struct llog_rec_hdr lsr_hdr;
3077 struct ost_id lsr_oi;
3083 struct llog_rec_tail lsr_tail;
3084 } __attribute__((packed));
3086 struct llog_size_change_rec {
3087 struct llog_rec_hdr lsc_hdr;
3088 struct ll_fid lsc_fid;
3093 struct llog_rec_tail lsc_tail;
3094 } __attribute__((packed));
3096 #define CHANGELOG_MAGIC 0xca103000
3098 /** \a changelog_rec_type's that can't be masked */
3099 #define CHANGELOG_MINMASK (1 << CL_MARK)
3100 /** bits covering all \a changelog_rec_type's */
3101 #define CHANGELOG_ALLMASK 0XFFFFFFFF
3102 /** default \a changelog_rec_type mask */
3103 #define CHANGELOG_DEFMASK CHANGELOG_ALLMASK & ~(1 << CL_ATIME | 1 << CL_CLOSE)
3105 /* changelog llog name, needed by client replicators */
3106 #define CHANGELOG_CATALOG "changelog_catalog"
3108 struct changelog_setinfo {
3111 } __attribute__((packed));
3113 /** changelog record */
3114 struct llog_changelog_rec {
3115 struct llog_rec_hdr cr_hdr;
3116 struct changelog_rec cr;
3117 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
3118 } __attribute__((packed));
3120 struct llog_changelog_ext_rec {
3121 struct llog_rec_hdr cr_hdr;
3122 struct changelog_ext_rec cr;
3123 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
3124 } __attribute__((packed));
3126 #define CHANGELOG_USER_PREFIX "cl"
3128 struct llog_changelog_user_rec {
3129 struct llog_rec_hdr cur_hdr;
3133 struct llog_rec_tail cur_tail;
3134 } __attribute__((packed));
3136 enum agent_req_status {
3144 static inline char *agent_req_status2name(enum agent_req_status ars)
3162 static inline bool agent_req_in_final_state(enum agent_req_status ars)
3164 return ((ars == ARS_SUCCEED) || (ars == ARS_FAILED) ||
3165 (ars == ARS_CANCELED));
3168 struct llog_agent_req_rec {
3169 struct llog_rec_hdr arr_hdr; /**< record header */
3170 __u32 arr_status; /**< status of the request */
3172 * agent_req_status */
3173 __u32 arr_archive_id; /**< backend archive number */
3174 __u64 arr_flags; /**< req flags */
3175 __u64 arr_compound_id; /**< compound cookie */
3176 __u64 arr_req_create; /**< req. creation time */
3177 __u64 arr_req_change; /**< req. status change time */
3178 struct hsm_action_item arr_hai; /**< req. to the agent */
3179 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
3180 } __attribute__((packed));
3182 /* Old llog gen for compatibility */
3186 } __attribute__((packed));
3188 struct llog_gen_rec {
3189 struct llog_rec_hdr lgr_hdr;
3190 struct llog_gen lgr_gen;
3194 struct llog_rec_tail lgr_tail;
3197 /* On-disk header structure of each log object, stored in little endian order */
3198 #define LLOG_CHUNK_SIZE 8192
3199 #define LLOG_HEADER_SIZE (96)
3200 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
3202 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
3204 /* flags for the logs */
3206 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
3207 LLOG_F_IS_CAT = 0x2,
3208 LLOG_F_IS_PLAIN = 0x4,
3211 struct llog_log_hdr {
3212 struct llog_rec_hdr llh_hdr;
3213 obd_time llh_timestamp;
3215 __u32 llh_bitmap_offset;
3219 /* for a catalog the first plain slot is next to it */
3220 struct obd_uuid llh_tgtuuid;
3221 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
3222 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
3223 struct llog_rec_tail llh_tail;
3224 } __attribute__((packed));
3226 #define LLOG_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
3227 llh->llh_bitmap_offset - \
3228 sizeof(llh->llh_tail)) * 8)
3230 /** log cookies are used to reference a specific log file and a record therein */
3231 struct llog_cookie {
3232 struct llog_logid lgc_lgl;
3236 } __attribute__((packed));
3238 /** llog protocol */
3239 enum llogd_rpc_ops {
3240 LLOG_ORIGIN_HANDLE_CREATE = 501,
3241 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
3242 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
3243 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
3244 LLOG_ORIGIN_HANDLE_CLOSE = 505,
3245 LLOG_ORIGIN_CONNECT = 506,
3246 LLOG_CATINFO = 507, /* deprecated */
3247 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
3248 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
3250 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
3254 struct llog_logid lgd_logid;
3256 __u32 lgd_llh_flags;
3258 __u32 lgd_saved_index;
3260 __u64 lgd_cur_offset;
3261 } __attribute__((packed));
3263 struct llogd_conn_body {
3264 struct llog_gen lgdc_gen;
3265 struct llog_logid lgdc_logid;
3266 __u32 lgdc_ctxt_idx;
3267 } __attribute__((packed));
3269 /* Note: 64-bit types are 64-bit aligned in structure */
3271 obd_valid o_valid; /* hot fields in this obdo */
3273 obd_id o_parent_seq;
3274 obd_size o_size; /* o_size-o_blocks == ost_lvb */
3278 obd_blocks o_blocks; /* brw: cli sent cached bytes */
3281 /* 32-bit fields start here: keep an even number of them via padding */
3282 obd_blksize o_blksize; /* optimal IO blocksize */
3283 obd_mode o_mode; /* brw: cli sent cache remain */
3287 obd_count o_nlink; /* brw: checksum */
3288 obd_count o_parent_oid;
3289 obd_count o_misc; /* brw: o_dropped */
3291 __u64 o_ioepoch; /* epoch in ost writes */
3292 __u32 o_stripe_idx; /* holds stripe idx */
3294 struct lustre_handle o_handle; /* brw: lock handle to prolong
3296 struct llog_cookie o_lcookie; /* destroy: unlink cookie from
3301 __u64 o_data_version; /* getattr: sum of iversion for
3303 * brw: grant space consumed on
3304 * the client for the write */
3310 #define o_dirty o_blocks
3311 #define o_undirty o_mode
3312 #define o_dropped o_misc
3313 #define o_cksum o_nlink
3314 #define o_grant_used o_data_version
3316 static inline void lustre_set_wire_obdo(struct obd_connect_data *ocd,
3317 struct obdo *wobdo, struct obdo *lobdo)
3319 memcpy(wobdo, lobdo, sizeof(*lobdo));
3320 wobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3324 if (unlikely(!(ocd->ocd_connect_flags & OBD_CONNECT_FID)) &&
3325 fid_seq_is_echo(ostid_seq(&lobdo->o_oi))) {
3326 /* Currently OBD_FL_OSTID will only be used when 2.4 echo
3327 * client communicate with pre-2.4 server */
3328 wobdo->o_oi.oi.oi_id = fid_oid(&lobdo->o_oi.oi_fid);
3329 wobdo->o_oi.oi.oi_seq = fid_seq(&lobdo->o_oi.oi_fid);
3333 static inline void lustre_get_wire_obdo(struct obd_connect_data *ocd,
3334 struct obdo *lobdo, struct obdo *wobdo)
3336 obd_flag local_flags = 0;
3338 if (lobdo->o_valid & OBD_MD_FLFLAGS)
3339 local_flags = lobdo->o_flags & OBD_FL_LOCAL_MASK;
3341 LASSERT(!(wobdo->o_flags & OBD_FL_LOCAL_MASK));
3343 memcpy(lobdo, wobdo, sizeof(*lobdo));
3344 if (local_flags != 0) {
3345 lobdo->o_valid |= OBD_MD_FLFLAGS;
3346 lobdo->o_flags &= ~OBD_FL_LOCAL_MASK;
3347 lobdo->o_flags |= local_flags;
3352 if (unlikely(!(ocd->ocd_connect_flags & OBD_CONNECT_FID)) &&
3353 fid_seq_is_echo(wobdo->o_oi.oi.oi_seq)) {
3355 lobdo->o_oi.oi_fid.f_seq = wobdo->o_oi.oi.oi_seq;
3356 lobdo->o_oi.oi_fid.f_oid = wobdo->o_oi.oi.oi_id;
3357 lobdo->o_oi.oi_fid.f_ver = 0;
3361 extern void lustre_swab_obdo (struct obdo *o);
3363 /* request structure for OST's */
3368 /* Key for FIEMAP to be used in get_info calls */
3369 struct ll_fiemap_info_key {
3372 struct ll_user_fiemap fiemap;
3375 extern void lustre_swab_ost_body (struct ost_body *b);
3376 extern void lustre_swab_ost_last_id(obd_id *id);
3377 extern void lustre_swab_fiemap(struct ll_user_fiemap *fiemap);
3379 extern void lustre_swab_lov_user_md_v1(struct lov_user_md_v1 *lum);
3380 extern void lustre_swab_lov_user_md_v3(struct lov_user_md_v3 *lum);
3381 extern void lustre_swab_lov_user_md_objects(struct lov_user_ost_data *lod,
3383 extern void lustre_swab_lov_mds_md(struct lov_mds_md *lmm);
3386 extern void lustre_swab_llogd_body (struct llogd_body *d);
3387 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
3388 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
3389 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec);
3390 extern void lustre_swab_llog_id(struct llog_logid *lid);
3393 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
3395 /* Functions for dumping PTLRPC fields */
3396 void dump_rniobuf(struct niobuf_remote *rnb);
3397 void dump_ioo(struct obd_ioobj *nb);
3398 void dump_obdo(struct obdo *oa);
3399 void dump_ost_body(struct ost_body *ob);
3400 void dump_rcs(__u32 *rc);
3402 #define IDX_INFO_MAGIC 0x3D37CC37
3404 /* Index file transfer through the network. The server serializes the index into
3405 * a byte stream which is sent to the client via a bulk transfer */
3409 /* reply: see idx_info_flags below */
3412 /* request & reply: number of lu_idxpage (to be) transferred */
3416 /* request: requested attributes passed down to the iterator API */
3419 /* request & reply: index file identifier (FID) */
3420 struct lu_fid ii_fid;
3422 /* reply: version of the index file before starting to walk the index.
3423 * Please note that the version can be modified at any time during the
3427 /* request: hash to start with:
3428 * reply: hash of the first entry of the first lu_idxpage and hash
3429 * of the entry to read next if any */
3430 __u64 ii_hash_start;
3433 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3437 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3445 extern void lustre_swab_idx_info(struct idx_info *ii);
3447 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3449 /* List of flags used in idx_info::ii_flags */
3450 enum idx_info_flags {
3451 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3452 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3453 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3454 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3457 #define LIP_MAGIC 0x8A6D6B6C
3459 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3461 /* 16-byte header */
3464 __u16 lip_nr; /* number of entries in the container */
3465 __u64 lip_pad0; /* additional padding for future use */
3467 /* key/record pairs are stored in the remaining 4080 bytes.
3468 * depending upon the flags in idx_info::ii_flags, each key/record
3469 * pair might be preceded by:
3471 * - the key size (II_FL_VARKEY is set)
3472 * - the record size (II_FL_VARREC is set)
3474 * For the time being, we only support fixed-size key & record. */
3475 char lip_entries[0];
3477 extern void lustre_swab_lip_header(struct lu_idxpage *lip);
3479 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3481 /* Gather all possible type associated with a 4KB container */
3483 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3484 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3485 char lp_array[LU_PAGE_SIZE];
3488 /* security opcodes */
3491 SEC_CTX_INIT_CONT = 802,
3494 SEC_FIRST_OPC = SEC_CTX_INIT
3498 * capa related definitions
3500 #define CAPA_HMAC_MAX_LEN 64
3501 #define CAPA_HMAC_KEY_MAX_LEN 56
3503 /* NB take care when changing the sequence of elements this struct,
3504 * because the offset info is used in find_capa() */
3505 struct lustre_capa {
3506 struct lu_fid lc_fid; /** fid */
3507 __u64 lc_opc; /** operations allowed */
3508 __u64 lc_uid; /** file owner */
3509 __u64 lc_gid; /** file group */
3510 __u32 lc_flags; /** HMAC algorithm & flags */
3511 __u32 lc_keyid; /** key# used for the capability */
3512 __u32 lc_timeout; /** capa timeout value (sec) */
3513 __u32 lc_expiry; /** expiry time (sec) */
3514 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
3515 } __attribute__((packed));
3517 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
3519 /** lustre_capa::lc_opc */
3521 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
3522 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
3523 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
3524 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
3525 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
3526 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
3527 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
3528 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
3529 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
3530 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
3531 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
3534 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3535 #define CAPA_OPC_MDS_ONLY \
3536 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3537 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3538 #define CAPA_OPC_OSS_ONLY \
3539 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3540 CAPA_OPC_OSS_DESTROY)
3541 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3542 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3544 /* MDS capability covers object capability for operations of body r/w
3545 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
3546 * while OSS capability only covers object capability for operations of
3547 * oss data(file content) r/w/truncate.
3549 static inline int capa_for_mds(struct lustre_capa *c)
3551 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
3554 static inline int capa_for_oss(struct lustre_capa *c)
3556 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
3559 /* lustre_capa::lc_hmac_alg */
3561 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
3565 #define CAPA_FL_MASK 0x00ffffff
3566 #define CAPA_HMAC_ALG_MASK 0xff000000
3568 struct lustre_capa_key {
3569 __u64 lk_seq; /**< mds# */
3570 __u32 lk_keyid; /**< key# */
3572 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
3573 } __attribute__((packed));
3575 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
3577 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3578 #define LINK_EA_MAGIC 0x11EAF1DFUL
3579 struct link_ea_header {
3582 __u64 leh_len; /* total size */
3588 /** Hardlink data is name and parent fid.
3589 * Stored in this crazy struct for maximum packing and endian-neutrality
3591 struct link_ea_entry {
3592 /** __u16 stored big-endian, unaligned */
3593 unsigned char lee_reclen[2];
3594 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3596 }__attribute__((packed));
3598 /** fid2path request/reply structure */
3599 struct getinfo_fid2path {
3600 struct lu_fid gf_fid;
3605 } __attribute__((packed));
3607 void lustre_swab_fid2path (struct getinfo_fid2path *gf);
3610 LAYOUT_INTENT_ACCESS = 0,
3611 LAYOUT_INTENT_READ = 1,
3612 LAYOUT_INTENT_WRITE = 2,
3613 LAYOUT_INTENT_GLIMPSE = 3,
3614 LAYOUT_INTENT_TRUNC = 4,
3615 LAYOUT_INTENT_RELEASE = 5,
3616 LAYOUT_INTENT_RESTORE = 6
3619 /* enqueue layout lock with intent */
3620 struct layout_intent {
3621 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3627 void lustre_swab_layout_intent(struct layout_intent *li);
3630 * On the wire version of hsm_progress structure.
3632 * Contains the userspace hsm_progress and some internal fields.
3634 struct hsm_progress_kernel {
3635 /* Field taken from struct hsm_progress */
3638 struct hsm_extent hpk_extent;
3640 __u16 hpk_errval; /* positive val */
3642 /* Additional fields */
3643 __u64 hpk_data_version;
3645 } __attribute__((packed));
3647 extern void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
3648 extern void lustre_swab_hsm_current_action(struct hsm_current_action *action);
3649 extern void lustre_swab_hsm_progress_kernel(struct hsm_progress_kernel *hpk);
3650 extern void lustre_swab_hsm_user_state(struct hsm_user_state *hus);
3651 extern void lustre_swab_hsm_user_item(struct hsm_user_item *hui);
3652 extern void lustre_swab_hsm_request(struct hsm_request *hr);
3655 * These are object update opcode under UPDATE_OBJ, which is currently
3656 * being used by cross-ref operations between MDT.
3658 * During the cross-ref operation, the Master MDT, which the client send the
3659 * request to, will disassembly the operation into object updates, then OSP
3660 * will send these updates to the remote MDT to be executed.
3662 * Update request format
3663 * magic: UPDATE_BUFFER_MAGIC_V1
3664 * Count: How many updates in the req.
3665 * bufs[0] : following are packets of object.
3667 * type: object_update_op, the op code of update
3668 * fid: The object fid of the update.
3669 * lens/bufs: other parameters of the update.
3671 * type: object_update_op, the op code of update
3672 * fid: The object fid of the update.
3673 * lens/bufs: other parameters of the update.
3675 * update[7]: type: object_update_op, the op code of update
3676 * fid: The object fid of the update.
3677 * lens/bufs: other parameters of the update.
3678 * Current 8 maxim updates per object update request.
3680 *******************************************************************
3681 * update reply format:
3683 * ur_version: UPDATE_REPLY_V1
3684 * ur_count: The count of the reply, which is usually equal
3685 * to the number of updates in the request.
3686 * ur_lens: The reply lengths of each object update.
3688 * replies: 1st update reply [4bytes_ret: other body]
3689 * 2nd update reply [4bytes_ret: other body]
3691 * nth update reply [4bytes_ret: other body]
3693 * For each reply of the update, the format would be
3694 * result(4 bytes):Other stuff
3697 #define UPDATE_MAX_OPS 10
3698 #define UPDATE_BUFFER_MAGIC_V1 0xBDDE0001
3699 #define UPDATE_BUFFER_MAGIC UPDATE_BUFFER_MAGIC_V1
3700 #define UPDATE_BUF_COUNT 8
3701 enum object_update_op {
3710 OBJ_INDEX_LOOKUP = 9,
3711 OBJ_INDEX_INSERT = 10,
3712 OBJ_INDEX_DELETE = 11,
3719 struct lu_fid u_fid;
3720 __u32 u_lens[UPDATE_BUF_COUNT];
3730 #define UPDATE_REPLY_V1 0x00BD0001
3731 struct update_reply {
3737 void lustre_swab_update_buf(struct update_buf *ub);
3738 void lustre_swab_update_reply_buf(struct update_reply *ur);
3740 /** layout swap request structure
3741 * fid1 and fid2 are in mdt_body
3743 struct mdc_swap_layouts {
3747 void lustre_swab_swap_layouts(struct mdc_swap_layouts *msl);
3750 struct lustre_handle cd_handle;
3751 struct lu_fid cd_fid;
3752 __u64 cd_data_version;
3753 __u64 cd_reserved[8];
3756 void lustre_swab_close_data(struct close_data *data);