1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
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 * We assume all nodes are either little-endian or big-endian, and we
46 * always send messages in the sender's native format. The receiver
47 * detects the message format by checking the 'magic' field of the message
48 * (see lustre_msg_swabbed() below).
49 * ALL structs passing over the wire should be declared here. Structs
50 * that are used in interfaces with userspace should go in lustre_user.h.
52 * All structs being declared here should be built from simple fixed-size
53 * types (__u8, __u16, __u32, __u64) or be built from other types or
54 * structs also declared in this file. Similarly, all flags and magic
55 * values in those structs should also be declared here. This ensures
56 * that the Lustre wire protocol is not influenced by external dependencies.
58 * The only other acceptable items in this file are VERY SIMPLE accessor
59 * functions to avoid callers grubbing inside the structures, and the
60 * prototypes of the swabber functions for each struct. Nothing that
61 * depends on external functions or definitions should be in here.
63 * Structs must be properly aligned to put 64-bit values on an 8-byte
64 * boundary. Any structs being added here must also be added to
65 * utils/wirecheck.c and "make newwiretest" run to regenerate the
66 * utils/wiretest.c sources. This allows us to verify that wire structs
67 * have the proper alignment/size on all architectures.
69 * DO NOT CHANGE any of the structs, flags, values declared here and used
70 * in released Lustre versions. Some structs may have padding fields that
71 * can be used. Some structs might allow addition at the end (verify this
72 * in the code to ensure that new/old clients that see this larger struct
73 * do not fail, otherwise you need to implement protocol compatibility).
75 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
76 * implemented either here, inline (trivial implementations) or in
77 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
78 * endian, in-place in the message buffer.
80 * A swabber takes a single pointer argument. The caller must already have
81 * verified that the length of the message buffer >= sizeof (type).
83 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
84 * may be defined that swabs just the variable part, after the caller has
85 * verified that the message buffer is large enough.
90 #ifndef _LUSTRE_IDL_H_
91 #define _LUSTRE_IDL_H_
93 #include <libcfs/libcfs.h>
95 /* Defn's shared with user-space. */
96 #include <lustre/lustre_user.h>
97 #include <lustre/ll_fiemap.h>
102 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
103 * FOO_REPLY_PORTAL is for incoming replies on the FOO
104 * FOO_BULK_PORTAL is for incoming bulk on the FOO
107 #define CONNMGR_REQUEST_PORTAL 1
108 #define CONNMGR_REPLY_PORTAL 2
109 //#define OSC_REQUEST_PORTAL 3
110 #define OSC_REPLY_PORTAL 4
111 //#define OSC_BULK_PORTAL 5
112 #define OST_IO_PORTAL 6
113 #define OST_CREATE_PORTAL 7
114 #define OST_BULK_PORTAL 8
115 //#define MDC_REQUEST_PORTAL 9
116 #define MDC_REPLY_PORTAL 10
117 //#define MDC_BULK_PORTAL 11
118 #define MDS_REQUEST_PORTAL 12
119 //#define MDS_REPLY_PORTAL 13
120 #define MDS_BULK_PORTAL 14
121 #define LDLM_CB_REQUEST_PORTAL 15
122 #define LDLM_CB_REPLY_PORTAL 16
123 #define LDLM_CANCEL_REQUEST_PORTAL 17
124 #define LDLM_CANCEL_REPLY_PORTAL 18
125 //#define PTLBD_REQUEST_PORTAL 19
126 //#define PTLBD_REPLY_PORTAL 20
127 //#define PTLBD_BULK_PORTAL 21
128 #define MDS_SETATTR_PORTAL 22
129 #define MDS_READPAGE_PORTAL 23
130 #define MDS_MDS_PORTAL 24
132 #define MGC_REPLY_PORTAL 25
133 #define MGS_REQUEST_PORTAL 26
134 #define MGS_REPLY_PORTAL 27
135 #define OST_REQUEST_PORTAL 28
136 #define FLD_REQUEST_PORTAL 29
137 #define SEQ_METADATA_PORTAL 30
138 #define SEQ_DATA_PORTAL 31
139 #define SEQ_CONTROLLER_PORTAL 32
144 #define SVC_RUNNING 8
145 #define SVC_STOPPING 16
146 #define SVC_STOPPED 32
149 #define PTL_RPC_MSG_REQUEST 4711
150 #define PTL_RPC_MSG_ERR 4712
151 #define PTL_RPC_MSG_REPLY 4713
153 /* DON'T use swabbed values of MAGIC as magic! */
154 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
155 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
157 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
158 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
160 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
162 #define PTLRPC_MSG_VERSION 0x00000003
163 #define LUSTRE_VERSION_MASK 0xffff0000
164 #define LUSTRE_OBD_VERSION 0x00010000
165 #define LUSTRE_MDS_VERSION 0x00020000
166 #define LUSTRE_OST_VERSION 0x00030000
167 #define LUSTRE_DLM_VERSION 0x00040000
168 #define LUSTRE_LOG_VERSION 0x00050000
169 #define LUSTRE_MGS_VERSION 0x00060000
171 typedef __u32 mdsno_t;
172 typedef __u64 seqno_t;
175 * Describes a range of sequence, lsr_start is included but lsr_end is
177 * Same structure is used in fld module where lsr_mdt field holds mdt id
181 struct lu_seq_range {
189 * returns width of given range \a r
192 static inline __u64 range_space(const struct lu_seq_range *range)
194 return range->lsr_end - range->lsr_start;
198 * initialize range to zero
201 static inline void range_init(struct lu_seq_range *range)
203 range->lsr_start = range->lsr_end = range->lsr_mdt = 0;
207 * check if given seq id \a s is within given range \a r
210 static inline int range_within(const struct lu_seq_range *range,
213 return s >= range->lsr_start && s < range->lsr_end;
217 * allocate \a w units of sequence from range \a from.
219 static inline void range_alloc(struct lu_seq_range *to,
220 struct lu_seq_range *from,
223 to->lsr_start = from->lsr_start;
224 to->lsr_end = from->lsr_start + width;
225 from->lsr_start += width;
228 static inline int range_is_sane(const struct lu_seq_range *range)
230 return (range->lsr_end >= range->lsr_start);
233 static inline int range_is_zero(const struct lu_seq_range *range)
235 return (range->lsr_start == 0 && range->lsr_end == 0);
238 static inline int range_is_exhausted(const struct lu_seq_range *range)
241 return range_space(range) == 0;
244 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x):%x"
246 #define PRANGE(range) \
247 (range)->lsr_start, \
251 /** \defgroup lu_fid lu_fid
257 * Fid is a cluster-wide unique identifier of a file or an object
258 * (stripe). Fids are never reused. Fids are transmitted across network (in
259 * the sender byte-ordering), and stored on disk in a packed form (struct
260 * lu_fid_pack) in a big-endian order.
264 * fid sequence. Sequence is a unit of migration: all files (objects)
265 * with fids from a given sequence are stored on the same
268 * Lustre should support 2 ^ 64 objects, thus even if one
269 * sequence has one object we will never reach this value.
272 /** fid number within sequence. */
275 * fid version, used to distinguish different versions (in the sense
276 * of snapshots, etc.) of the same file system object. Not currently
283 * Following struct for MDT attributes, that will be kept inode's EA.
284 * Introduced in 2.0 release (please see b15993, for details)
286 struct lustre_mdt_attrs {
287 /** FID of this inode */
288 struct lu_fid lma_self_fid;
289 /** SOM state, mdt/ost type, others */
291 /** total sectors in objects */
292 __u64 lma_som_sectors;
300 /** initial fid id value */
301 LUSTRE_FID_INIT_OID = 1UL
304 /** returns fid object sequence */
305 static inline __u64 fid_seq(const struct lu_fid *fid)
310 /** returns fid object id */
311 static inline __u32 fid_oid(const struct lu_fid *fid)
316 /** returns fid object version */
317 static inline __u32 fid_ver(const struct lu_fid *fid)
322 static inline void fid_zero(struct lu_fid *fid)
324 memset(fid, 0, sizeof(*fid));
327 /* Normal FID sequence starts from this value, i.e. 1<<33 */
328 #define FID_SEQ_START 0x200000000ULL
330 /* IDIF sequence starts from this value, i.e. 1<<32 */
331 #define IDIF_SEQ_START 0x100000000ULL
334 * Check if a fid is igif or not.
335 * \param fid the fid to be tested.
336 * \return true if the fid is a igif; otherwise false.
338 static inline int fid_is_igif(const struct lu_fid *fid)
340 return fid_seq(fid) > 0 && fid_seq(fid) < IDIF_SEQ_START;
344 * Check if a fid is idif or not.
345 * \param fid the fid to be tested.
346 * \return true if the fid is a idif; otherwise false.
348 static inline int fid_is_idif(const struct lu_fid *fid)
350 return fid_seq(fid) >= IDIF_SEQ_START && fid_seq(fid) < FID_SEQ_START;
354 * Get inode number from a igif.
355 * \param fid a igif to get inode number from.
356 * \return inode number for the igif.
358 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
364 * Get inode generation from a igif.
365 * \param fid a igif to get inode generation from.
366 * \return inode generation for the igif.
368 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
373 #define DFID "["LPX64":0x%x:0x%x]"
380 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
382 /* check that all fields are converted */
383 CLASSERT(sizeof *src ==
384 sizeof fid_seq(src) +
385 sizeof fid_oid(src) + sizeof fid_ver(src));
386 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
387 dst->f_seq = cpu_to_le64(fid_seq(src));
388 dst->f_oid = cpu_to_le32(fid_oid(src));
389 dst->f_ver = cpu_to_le32(fid_ver(src));
392 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
394 /* check that all fields are converted */
395 CLASSERT(sizeof *src ==
396 sizeof fid_seq(src) +
397 sizeof fid_oid(src) + sizeof fid_ver(src));
398 dst->f_seq = le64_to_cpu(fid_seq(src));
399 dst->f_oid = le32_to_cpu(fid_oid(src));
400 dst->f_ver = le32_to_cpu(fid_ver(src));
401 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
404 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
406 /* check that all fields are converted */
407 CLASSERT(sizeof *src ==
408 sizeof fid_seq(src) +
409 sizeof fid_oid(src) + sizeof fid_ver(src));
410 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
411 dst->f_seq = cpu_to_be64(fid_seq(src));
412 dst->f_oid = cpu_to_be32(fid_oid(src));
413 dst->f_ver = cpu_to_be32(fid_ver(src));
416 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
418 /* check that all fields are converted */
419 CLASSERT(sizeof *src ==
420 sizeof fid_seq(src) +
421 sizeof fid_oid(src) + sizeof fid_ver(src));
422 dst->f_seq = be64_to_cpu(fid_seq(src));
423 dst->f_oid = be32_to_cpu(fid_oid(src));
424 dst->f_ver = be32_to_cpu(fid_ver(src));
425 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
429 * Storage representation for fids.
431 * Variable size, first byte contains the length of the whole record.
435 char fp_area[sizeof(struct lu_fid)];
438 void fid_pack(struct lu_fid_pack *pack, const struct lu_fid *fid,
439 struct lu_fid *befider);
440 int fid_unpack(const struct lu_fid_pack *pack, struct lu_fid *fid);
442 static inline int fid_is_sane(const struct lu_fid *fid)
446 ((fid_seq(fid) >= FID_SEQ_START && fid_oid(fid) != 0
447 && fid_ver(fid) == 0) ||
451 static inline int fid_is_zero(const struct lu_fid *fid)
453 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
456 extern void lustre_swab_lu_fid(struct lu_fid *fid);
457 extern void lustre_swab_lu_seq_range(struct lu_seq_range *range);
459 static inline int lu_fid_eq(const struct lu_fid *f0,
460 const struct lu_fid *f1)
462 /* Check that there is no alignment padding. */
463 CLASSERT(sizeof *f0 ==
464 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
465 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
466 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
467 return memcmp(f0, f1, sizeof *f0) == 0;
470 #define __diff_normalize(val0, val1) \
472 typeof(val0) __val0 = (val0); \
473 typeof(val1) __val1 = (val1); \
475 (__val0 == __val1 ? 0 : __val0 > __val1 ? +1 : -1); \
478 static inline int lu_fid_cmp(const struct lu_fid *f0,
479 const struct lu_fid *f1)
482 __diff_normalize(fid_seq(f0), fid_seq(f1)) ?:
483 __diff_normalize(fid_oid(f0), fid_oid(f1)) ?:
484 __diff_normalize(fid_ver(f0), fid_ver(f1));
489 /** \defgroup lu_dir lu_dir
492 * Layout of readdir pages, as transmitted on wire.
495 struct lu_fid lde_fid;
504 __u64 ldp_hash_start;
508 struct lu_dirent ldp_entries[0];
511 enum lu_dirpage_flags {
515 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
517 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
520 return dp->ldp_entries;
523 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
525 struct lu_dirent *next;
527 if (le16_to_cpu(ent->lde_reclen) != 0)
528 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
535 static inline int lu_dirent_size(struct lu_dirent *ent)
537 if (le16_to_cpu(ent->lde_reclen) == 0) {
538 return (sizeof(*ent) +
539 le16_to_cpu(ent->lde_namelen) + 7) & ~7;
541 return le16_to_cpu(ent->lde_reclen);
544 #define DIR_END_OFF 0xfffffffffffffffeULL
548 struct lustre_handle {
551 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
553 static inline int lustre_handle_is_used(struct lustre_handle *lh)
555 return lh->cookie != 0ull;
558 static inline int lustre_handle_equal(struct lustre_handle *lh1,
559 struct lustre_handle *lh2)
561 return lh1->cookie == lh2->cookie;
564 static inline void lustre_handle_copy(struct lustre_handle *tgt,
565 struct lustre_handle *src)
567 tgt->cookie = src->cookie;
570 /* flags for lm_flags */
571 #define MSGHDR_AT_SUPPORT 0x1
573 #define lustre_msg lustre_msg_v2
574 /* we depend on this structure to be 8-byte aligned */
575 /* this type is only endian-adjusted in lustre_unpack_msg() */
576 struct lustre_msg_v2 {
588 /* without gss, ptlrpc_body is put at the first buffer. */
589 #define PTLRPC_NUM_VERSIONS 4
591 struct lustre_handle pb_handle;
598 __u64 pb_last_committed;
603 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
604 __u32 pb_service_time; /* for rep, actual service time */
607 /* VBR: pre-versions */
608 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
609 /* padding for future needs */
613 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
615 /* message body offset for lustre_msg_v2 */
616 /* ptlrpc body offset in all request/reply messages */
617 #define MSG_PTLRPC_BODY_OFF 0
619 /* normal request/reply message record offset */
620 #define REQ_REC_OFF 1
621 #define REPLY_REC_OFF 1
623 /* ldlm request message body offset */
624 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
625 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
627 /* ldlm intent lock message body offset */
628 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
629 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
631 /* ldlm reply message body offset */
632 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
633 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
635 /* Flags that are operation-specific go in the top 16 bits. */
636 #define MSG_OP_FLAG_MASK 0xffff0000
637 #define MSG_OP_FLAG_SHIFT 16
639 /* Flags that apply to all requests are in the bottom 16 bits */
640 #define MSG_GEN_FLAG_MASK 0x0000ffff
641 #define MSG_LAST_REPLAY 0x0001
642 #define MSG_RESENT 0x0002
643 #define MSG_REPLAY 0x0004
644 /* #define MSG_AT_SUPPORT 0x0008
645 * This was used in early prototypes of adaptive timeouts, and while there
646 * shouldn't be any users of that code there also isn't a need for using this
647 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
648 #define MSG_DELAY_REPLAY 0x0010
649 #define MSG_VERSION_REPLAY 0x0020
650 #define MSG_REQ_REPLAY_DONE 0x0040
651 #define MSG_LOCK_REPLAY_DONE 0x0080
654 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
657 #define MSG_CONNECT_RECOVERING 0x1
658 #define MSG_CONNECT_RECONNECT 0x2
659 #define MSG_CONNECT_REPLAYABLE 0x4
660 //#define MSG_CONNECT_PEER 0x8
661 #define MSG_CONNECT_LIBCLIENT 0x10
662 #define MSG_CONNECT_INITIAL 0x20
663 #define MSG_CONNECT_ASYNC 0x40
664 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
665 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
668 #define OBD_CONNECT_RDONLY 0x00000001ULL /* client allowed read-only access */
669 #define OBD_CONNECT_INDEX 0x00000002ULL /* connect to specific LOV idx */
670 #define OBD_CONNECT_GRANT 0x00000008ULL /* OSC acquires grant at connect */
671 #define OBD_CONNECT_SRVLOCK 0x00000010ULL /* server takes locks for client */
672 #define OBD_CONNECT_VERSION 0x00000020ULL /* Server supports versions in ocd */
673 #define OBD_CONNECT_REQPORTAL 0x00000040ULL /* Separate portal for non-IO reqs */
674 #define OBD_CONNECT_ACL 0x00000080ULL /* client uses access control lists */
675 #define OBD_CONNECT_XATTR 0x00000100ULL /* client using extended attributes*/
676 #define OBD_CONNECT_TRUNCLOCK 0x00000400ULL /* locks on server for punch b=9528 */
677 #define OBD_CONNECT_IBITS 0x00001000ULL /* support for inodebits locks */
678 #define OBD_CONNECT_JOIN 0x00002000ULL /* files can be concatenated */
679 #define OBD_CONNECT_ATTRFID 0x00004000ULL /* Server supports GetAttr By Fid */
680 #define OBD_CONNECT_NODEVOH 0x00008000ULL /* No open handle for special nodes */
681 #define OBD_CONNECT_RMT_CLIENT 0x00010000ULL /* Remote client */
682 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x00020000ULL /* Remote client by force */
683 #define OBD_CONNECT_BRW_SIZE 0x00040000ULL /* Max bytes per rpc */
684 #define OBD_CONNECT_QUOTA64 0x00080000ULL /* 64bit qunit_data.qd_count b=10707*/
685 #define OBD_CONNECT_MDS_CAPA 0x00100000ULL /* MDS capability */
686 #define OBD_CONNECT_OSS_CAPA 0x00200000ULL /* OSS capability */
687 #define OBD_CONNECT_CANCELSET 0x00400000ULL /* Early batched cancels. */
688 #define OBD_CONNECT_SOM 0x00800000ULL /* SOM feature */
689 #define OBD_CONNECT_AT 0x01000000ULL /* client uses adaptive timeouts */
690 #define OBD_CONNECT_LRU_RESIZE 0x02000000ULL /* Lru resize feature. */
691 #define OBD_CONNECT_MDS_MDS 0x04000000ULL /* MDS-MDS connection*/
692 #define OBD_CONNECT_REAL 0x08000000ULL /* real connection */
693 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /* shrink/enlarge qunit b=10600 */
694 #define OBD_CONNECT_CKSUM 0x20000000ULL /* support several cksum algos */
695 #define OBD_CONNECT_FID 0x40000000ULL /* FID is supported by server */
696 #define OBD_CONNECT_LOV_V3 0x100000000ULL /* client supports lov v3 ea */
698 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
699 * and lustre/utils/wirecheck.c */
701 #ifdef HAVE_LRU_RESIZE_SUPPORT
702 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
704 #define LRU_RESIZE_CONNECT_FLAG 0
707 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
708 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
709 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
710 OBD_CONNECT_NODEVOH |/* OBD_CONNECT_ATTRFID |*/\
711 OBD_CONNECT_RMT_CLIENT | \
712 OBD_CONNECT_RMT_CLIENT_FORCE | \
713 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | \
714 OBD_CONNECT_MDS_MDS | OBD_CONNECT_CANCELSET | \
716 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_AT | \
718 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
719 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
720 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
721 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
722 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET | \
723 OBD_CONNECT_CKSUM | LRU_RESIZE_CONNECT_FLAG | \
724 OBD_CONNECT_AT | OBD_CONNECT_CHANGE_QS | \
725 OBD_CONNECT_RMT_CLIENT | \
726 OBD_CONNECT_RMT_CLIENT_FORCE)
727 #define ECHO_CONNECT_SUPPORTED (0)
728 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT)
730 #define MAX_QUOTA_COUNT32 (0xffffffffULL)
732 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
733 ((patch)<<8) + (fix))
734 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
735 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
736 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
737 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
739 /* This structure is used for both request and reply.
741 * If we eventually have separate connect data for different types, which we
742 * almost certainly will, then perhaps we stick a union in here. */
743 struct obd_connect_data {
744 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
745 __u32 ocd_version; /* lustre release version number */
746 __u32 ocd_grant; /* initial cache grant amount (bytes) */
747 __u32 ocd_index; /* LOV index to connect to */
748 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
749 __u64 ocd_ibits_known; /* inode bits this client understands */
750 __u32 ocd_nllu; /* non-local-lustre-user */
751 __u32 ocd_nllg; /* non-local-lustre-group */
752 __u64 ocd_transno; /* first transno from client to be replayed */
753 __u32 ocd_group; /* MDS group on OST */
754 __u32 ocd_cksum_types; /* supported checksum algorithms */
755 __u64 padding1; /* also fix lustre_swab_connect */
756 __u64 padding2; /* also fix lustre_swab_connect */
759 extern void lustre_swab_connect(struct obd_connect_data *ocd);
762 * Supported checksum algorithms. Up to 32 checksum types are supported.
763 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
764 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
765 * algorithm and also the OBD_FL_CKSUM* flags.
768 OBD_CKSUM_CRC32 = 0x00000001,
769 OBD_CKSUM_ADLER = 0x00000002,
773 * OST requests: OBDO & OBD request records
778 OST_REPLY = 0, /* reply ? */
796 OST_QUOTA_ADJUST_QUNIT = 20,
799 #define OST_FIRST_OPC OST_REPLY
801 typedef __u64 obd_id;
802 typedef __u64 obd_gr;
803 typedef __u64 obd_time;
804 typedef __u64 obd_size;
805 typedef __u64 obd_off;
806 typedef __u64 obd_blocks;
807 typedef __u32 obd_blksize;
808 typedef __u32 obd_mode;
809 typedef __u32 obd_uid;
810 typedef __u32 obd_gid;
811 typedef __u32 obd_flag;
812 typedef __u64 obd_valid;
813 typedef __u32 obd_count;
815 #define OBD_FL_INLINEDATA (0x00000001)
816 #define OBD_FL_OBDMDEXISTS (0x00000002)
817 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
818 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
819 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
820 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
821 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
822 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
823 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
826 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
827 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
830 #define OBD_FL_TRUNCLOCK (0x00000800)
835 #define OBD_FL_CKSUM_CRC32 (0x00001000)
836 #define OBD_FL_CKSUM_ADLER (0x00002000)
837 #define OBD_FL_CKSUM_ALL (OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER)
839 #define LOV_MAGIC_V1 0x0BD10BD0
840 #define LOV_MAGIC LOV_MAGIC_V1
841 #define LOV_MAGIC_JOIN 0x0BD20BD0
842 #define LOV_MAGIC_V3 0x0BD30BD0
844 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
845 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
846 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
847 #define LOV_PATTERN_CMOBD 0x200
849 #define LOV_OBJECT_GROUP_DEFAULT ~0ULL
850 #define LOV_OBJECT_GROUP_CLEAR 0ULL
852 #define lov_ost_data lov_ost_data_v1
853 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
854 __u64 l_object_id; /* OST object ID */
855 __u64 l_object_gr; /* OST object group (creating MDS number) */
856 __u32 l_ost_gen; /* generation of this l_ost_idx */
857 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
860 #define lov_mds_md lov_mds_md_v1
861 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
862 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
863 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
864 __u64 lmm_object_id; /* LOV object ID */
865 __u64 lmm_object_gr; /* LOV object group */
866 __u32 lmm_stripe_size; /* size of stripe in bytes */
867 __u32 lmm_stripe_count; /* num stripes in use for this object */
868 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
871 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
873 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
874 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
876 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
877 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
878 #define XATTR_USER_PREFIX "user."
879 #define XATTR_TRUSTED_PREFIX "trusted."
880 #define XATTR_SECURITY_PREFIX "security."
881 #define XATTR_LUSTRE_PREFIX "lustre."
883 #define XATTR_NAME_LOV "trusted.lov"
884 #define XATTR_NAME_LMA "trusted.lma"
885 #define XATTR_NAME_LMV "trusted.lmv"
886 #define XATTR_NAME_LINK "trusted.link"
889 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
890 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
891 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
892 __u64 lmm_object_id; /* LOV object ID */
893 __u64 lmm_object_gr; /* LOV object group */
894 __u32 lmm_stripe_size; /* size of stripe in bytes */
895 __u32 lmm_stripe_count; /* num stripes in use for this object */
896 char lmm_pool_name[LOV_MAXPOOLNAME]; /* must be 32bit aligned */
897 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
901 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
902 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
903 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
904 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
905 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
906 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
907 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
908 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
909 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
910 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
911 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
912 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
913 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
914 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
915 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
916 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
917 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
918 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
919 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
920 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
921 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
922 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
923 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
924 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
925 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
926 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
927 /* ->mds if epoch opens or closes */
928 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
929 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
930 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
931 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
932 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
934 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
935 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
936 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
938 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
939 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
940 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
941 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
942 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
943 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
944 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
945 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
946 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
948 #define OBD_FL_TRUNC (0x0000200000000000ULL) /* for filter_truncate */
950 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
951 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
952 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
953 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
955 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
956 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
957 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
958 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
959 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
961 /* don't forget obdo_fid which is way down at the bottom so it can
962 * come after the definition of llog_cookie */
975 __u32 os_state; /* positive error code on server */
987 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
988 #define OBD_STATFS_NODELAY 0x0001 /* requests should be send without delay
989 * and resends for avoid deadlocks */
991 #define OBD_STATFS_FROM_CACHE 0x0002 /* the statfs callback should not update
994 /* ost_body.data values for OST_BRW */
996 #define OBD_BRW_READ 0x01
997 #define OBD_BRW_WRITE 0x02
998 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
999 #define OBD_BRW_SYNC 0x08
1000 #define OBD_BRW_CHECK 0x10
1001 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1002 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1003 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1004 #define OBD_BRW_NOQUOTA 0x100
1005 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1007 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
1009 #define OST_MIN_PRECREATE 32
1010 #define OST_MAX_PRECREATE 20000
1019 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
1021 /* multiple of 8 bytes => can array */
1022 struct niobuf_remote {
1028 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
1030 /* lock value block communicated between the filter and llite */
1032 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1033 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1034 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1035 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1036 #define OST_LVB_IS_ERR(blocks) \
1037 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1038 #define OST_LVB_SET_ERR(blocks, rc) \
1039 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1040 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1050 extern void lustre_swab_ost_lvb(struct ost_lvb *);
1059 MDS_GETATTR_NAME = 34,
1064 MDS_DISCONNECT = 39,
1070 MDS_DONE_WRITING = 45,
1072 MDS_QUOTACHECK = 47,
1075 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1081 #define MDS_FIRST_OPC MDS_GETATTR
1098 } mds_reint_t, mdt_reint_t;
1100 extern void lustre_swab_generic_32s (__u32 *val);
1102 /* the disposition of the intent outlines what was executed */
1103 #define DISP_IT_EXECD 0x00000001
1104 #define DISP_LOOKUP_EXECD 0x00000002
1105 #define DISP_LOOKUP_NEG 0x00000004
1106 #define DISP_LOOKUP_POS 0x00000008
1107 #define DISP_OPEN_CREATE 0x00000010
1108 #define DISP_OPEN_OPEN 0x00000020
1109 #define DISP_ENQ_COMPLETE 0x00400000
1110 #define DISP_ENQ_OPEN_REF 0x00800000
1111 #define DISP_ENQ_CREATE_REF 0x01000000
1112 #define DISP_OPEN_LOCK 0x02000000
1114 /* INODE LOCK PARTS */
1115 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1116 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1117 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1119 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
1120 #define MDS_INODELOCK_MAXSHIFT 2
1121 /* This FULL lock is useful to take on unlink sort of operations */
1122 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1125 __u64 id; /* holds object id */
1126 __u32 generation; /* holds object generation */
1128 __u32 f_type; /* holds object type or stripe idx when passing it to
1129 * OST for saving into EA. */
1132 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1134 #define MDS_STATUS_CONN 1
1135 #define MDS_STATUS_LOV 2
1137 struct mds_status_req {
1142 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
1144 /* mdt_thread_info.mti_flags. */
1146 /* The flag indicates Size-on-MDS attributes are changed. */
1147 MF_SOM_CHANGE = (1 << 0),
1148 /* Flags indicates an epoch opens or closes. */
1149 MF_EPOCH_OPEN = (1 << 1),
1150 MF_EPOCH_CLOSE = (1 << 2),
1151 MF_MDC_CANCEL_FID1 = (1 << 3),
1152 MF_MDC_CANCEL_FID2 = (1 << 4),
1153 MF_MDC_CANCEL_FID3 = (1 << 5),
1154 MF_MDC_CANCEL_FID4 = (1 << 6),
1157 #define MF_SOM_LOCAL_FLAGS (MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID2 | \
1158 MF_MDC_CANCEL_FID3 | MF_MDC_CANCEL_FID4)
1160 #define MDS_BFLAG_UNCOMMITTED_WRITES 0x1
1161 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
1163 /* these should be identical to their EXT3_*_FL counterparts, and are
1164 * redefined here only to avoid dragging in ext3_fs.h */
1165 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
1166 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
1167 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
1168 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
1169 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1172 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
1173 * and we need to decode these into local S_* flags in the inode. Otherwise
1174 * we pass flags straight through (see bug 9486). */
1175 static inline int ll_ext_to_inode_flags(int flags)
1177 return (flags & MDS_BFLAG_EXT_FLAGS) ?
1178 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
1179 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
1180 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
1181 #if defined(S_DIRSYNC)
1182 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1184 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1185 (flags & ~MDS_BFLAG_EXT_FLAGS);
1188 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1189 * and we pass these straight through. Otherwise we need to convert from
1190 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1191 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1193 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1194 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1195 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1196 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1197 #if defined(S_DIRSYNC)
1198 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1200 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1207 struct lustre_handle handle;
1209 __u64 size; /* Offset, in the case of MDS_READPAGE */
1213 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1215 __u64 ino; /* for 1.6 compatibility */
1222 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1224 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1225 __u32 generation; /* for 1.6 compatibility */
1230 __u32 max_cookiesize;
1231 __u32 padding_4; /* also fix lustre_swab_mdt_body */
1243 struct lustre_handle handle;
1245 __u64 size; /* Offset, in the case of MDS_READPAGE */
1249 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1258 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1260 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1266 __u32 max_cookiesize;
1267 __u32 padding_4; /* also fix lustre_swab_mds_body */
1270 extern void lustre_swab_mds_body (struct mds_body *b);
1271 extern void lustre_swab_mdt_body (struct mdt_body *b);
1274 struct lustre_handle handle;
1280 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1282 #define Q_QUOTACHECK 0x800100
1283 #define Q_INITQUOTA 0x800101 /* init slave limits */
1284 #define Q_GETOINFO 0x800102 /* get obd quota info */
1285 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1286 #define Q_FINVALIDATE 0x800104 /* invalidate operational quotas */
1288 #define Q_TYPEMATCH(id, type) \
1289 ((id) == (type) || (id) == UGQUOTA)
1291 #define Q_TYPESET(oqc, type) Q_TYPEMATCH((oqc)->qc_type, type)
1293 #define Q_GETOCMD(oqc) \
1294 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1296 #define QCTL_COPY(out, in) \
1298 Q_COPY(out, in, qc_cmd); \
1299 Q_COPY(out, in, qc_type); \
1300 Q_COPY(out, in, qc_id); \
1301 Q_COPY(out, in, qc_stat); \
1302 Q_COPY(out, in, qc_dqinfo); \
1303 Q_COPY(out, in, qc_dqblk); \
1306 struct obd_quotactl {
1311 struct obd_dqinfo qc_dqinfo;
1312 struct obd_dqblk qc_dqblk;
1315 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1317 struct quota_adjust_qunit {
1324 extern void lustre_swab_quota_adjust_qunit(struct quota_adjust_qunit *q);
1326 /* flags in qunit_data and quota_adjust_qunit will use macroes below */
1327 #define LQUOTA_FLAGS_GRP 1UL /* 0 is user, 1 is group */
1328 #define LQUOTA_FLAGS_BLK 2UL /* 0 is inode, 1 is block */
1329 #define LQUOTA_FLAGS_ADJBLK 4UL /* adjust the block qunit size */
1330 #define LQUOTA_FLAGS_ADJINO 8UL /* adjust the inode qunit size */
1331 #define LQUOTA_FLAGS_CHG_QS 16UL /* indicate whether it has capability of
1332 * OBD_CONNECT_CHANGE_QS */
1334 /* the status of lqsk_flags in struct lustre_qunit_size_key */
1335 #define LQUOTA_QUNIT_FLAGS (LQUOTA_FLAGS_GRP | LQUOTA_FLAGS_BLK)
1337 #define QAQ_IS_GRP(qaq) ((qaq)->qaq_flags & LQUOTA_FLAGS_GRP)
1338 #define QAQ_IS_ADJBLK(qaq) ((qaq)->qaq_flags & LQUOTA_FLAGS_ADJBLK)
1339 #define QAQ_IS_ADJINO(qaq) ((qaq)->qaq_flags & LQUOTA_FLAGS_ADJINO)
1341 #define QAQ_SET_GRP(qaq) ((qaq)->qaq_flags |= LQUOTA_FLAGS_GRP)
1342 #define QAQ_SET_ADJBLK(qaq) ((qaq)->qaq_flags |= LQUOTA_FLAGS_ADJBLK)
1343 #define QAQ_SET_ADJINO(qaq) ((qaq)->qaq_flags |= LQUOTA_FLAGS_ADJINO)
1345 /* inode access permission for remote user, the inode info are omitted,
1346 * for client knows them. */
1347 struct mds_remote_perm {
1352 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1355 /* permissions for md_perm.mp_perm */
1357 CFS_SETUID_PERM = 0x01,
1358 CFS_SETGID_PERM = 0x02,
1359 CFS_SETGRP_PERM = 0x04,
1360 CFS_RMTACL_PERM = 0x08,
1361 CFS_RMTOWN_PERM = 0x10
1364 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1366 struct mdt_remote_perm {
1371 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1374 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1376 struct mds_rec_setattr {
1383 struct ll_fid sa_fid;
1384 __u64 sa_valid; /* MDS_ATTR_* attributes */
1391 __u32 sa_attr_flags;
1392 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1395 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1397 struct mdt_rec_setattr {
1407 __u32 sa_padding_1_h;
1408 struct lu_fid sa_fid;
1417 __u32 sa_attr_flags;
1425 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1428 * Attribute flags used in mds_rec_setattr::sa_valid.
1429 * The kernel's #defines for ATTR_* should not be used over the network
1430 * since the client and MDS may run different kernels (see bug 13828)
1431 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1433 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1434 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1435 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1436 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1437 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1438 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1439 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1440 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1441 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1442 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1443 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1444 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1445 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1446 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1447 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1448 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1451 #define FMODE_READ 00000001
1452 #define FMODE_WRITE 00000002
1455 #define FMODE_EPOCH 01000000
1456 #define FMODE_EPOCHLCK 02000000
1457 #define FMODE_SOM 04000000
1458 #define FMODE_CLOSED 0
1460 #define MDS_OPEN_CREATED 00000010
1461 #define MDS_OPEN_CROSS 00000020
1463 #define MDS_FMODE_EXEC 00000004
1464 #define MDS_OPEN_CREAT 00000100
1465 #define MDS_OPEN_EXCL 00000200
1466 #define MDS_OPEN_TRUNC 00001000
1467 #define MDS_OPEN_APPEND 00002000
1468 #define MDS_OPEN_SYNC 00010000
1469 #define MDS_OPEN_DIRECTORY 00200000
1471 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1472 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1473 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1474 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1475 * with default ACL */
1476 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1477 * actually, this is for create, not
1478 * conflict with other open flags */
1479 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1480 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1481 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1483 /* permission for create non-directory file */
1484 #define MAY_CREATE (1 << 7)
1485 /* permission for create directory file */
1486 #define MAY_LINK (1 << 8)
1487 /* permission for delete from the directory */
1488 #define MAY_UNLINK (1 << 9)
1489 /* source's permission for rename */
1490 #define MAY_RENAME_SRC (1 << 10)
1491 /* target's permission for rename */
1492 #define MAY_RENAME_TAR (1 << 11)
1493 /* part (parent's) VTX permission check */
1494 #define MAY_VTX_PART (1 << 12)
1495 /* full VTX permission check */
1496 #define MAY_VTX_FULL (1 << 13)
1497 /* lfs rgetfacl permission check */
1498 #define MAY_RGETFACL (1 << 14)
1501 MDS_CHECK_SPLIT = 1 << 0,
1502 MDS_CROSS_REF = 1 << 1,
1503 MDS_VTX_BYPASS = 1 << 2,
1504 MDS_PERM_BYPASS = 1 << 3,
1506 MDS_QUOTA_IGNORE = 1 << 5
1509 struct mds_rec_join {
1510 struct ll_fid jr_fid;
1514 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1516 struct mdt_rec_join {
1517 struct lu_fid jr_fid;
1521 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1523 struct mds_rec_create {
1528 __u32 cr_flags; /* for use with open */
1530 struct ll_fid cr_fid;
1531 struct ll_fid cr_replayfid;
1535 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1536 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1537 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1538 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1539 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1542 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1544 struct mdt_rec_create {
1552 __u32 cr_suppgid1_h;
1554 __u32 cr_suppgid2_h;
1555 struct lu_fid cr_fid1;
1556 struct lu_fid cr_fid2;
1557 struct lustre_handle cr_old_handle; /* u64 handle in case of open replay */
1561 __u64 cr_padding_1; /* pad for 64 bits*/
1564 __u32 cr_flags; /* for use with open */
1570 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1572 struct mds_rec_link {
1579 struct ll_fid lk_fid1;
1580 struct ll_fid lk_fid2;
1582 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1583 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1584 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1585 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1588 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1590 struct mdt_rec_link {
1598 __u32 lk_suppgid1_h;
1600 __u32 lk_suppgid2_h;
1601 struct lu_fid lk_fid1;
1602 struct lu_fid lk_fid2;
1616 struct mds_rec_unlink {
1623 struct ll_fid ul_fid1;
1624 struct ll_fid ul_fid2;
1626 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1627 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1628 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1629 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1632 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1634 struct mdt_rec_unlink {
1642 __u32 ul_suppgid1_h;
1644 __u32 ul_suppgid2_h;
1645 struct lu_fid ul_fid1;
1646 struct lu_fid ul_fid2;
1660 struct mds_rec_rename {
1667 struct ll_fid rn_fid1;
1668 struct ll_fid rn_fid2;
1670 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1671 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1672 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1673 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1676 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1678 struct mdt_rec_rename {
1686 __u32 rn_suppgid1_h;
1688 __u32 rn_suppgid2_h;
1689 struct lu_fid rn_fid1;
1690 struct lu_fid rn_fid2;
1696 __u32 rn_bias; /* some operation flags */
1697 __u32 rn_mode; /* cross-ref rename has mode */
1704 struct mdt_rec_setxattr {
1712 __u32 sx_suppgid1_h;
1714 __u32 sx_suppgid2_h;
1715 struct lu_fid sx_fid;
1716 __u64 sx_padding_1; /* These three members are lu_fid size */
1728 __u32 sx_padding_10;
1729 __u32 sx_padding_11;
1732 struct mdt_rec_reint {
1740 __u32 rr_suppgid1_h;
1742 __u32 rr_suppgid2_h;
1743 struct lu_fid rr_fid1;
1744 struct lu_fid rr_fid2;
1752 __u32 rr_padding_1; /* also fix lustre_swab_mdt_rec_reint */
1753 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
1754 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
1755 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1758 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
1761 __u32 ld_tgt_count; /* how many MDS's */
1762 __u32 ld_active_tgt_count; /* how many active */
1763 struct obd_uuid ld_uuid;
1766 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1771 FLD_FIRST_OPC = FLD_QUERY
1777 SEQ_FIRST_OPC = SEQ_QUERY
1781 SEQ_ALLOC_SUPER = 0,
1786 * LOV data structures
1789 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1790 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1791 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
1793 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1794 /* The size of the buffer the lov/mdc reserves for the
1795 * array of UUIDs returned by the MDS. With the current
1796 * protocol, this will limit the max number of OSTs per LOV */
1798 #define LOV_DESC_MAGIC 0xB0CCDE5C
1800 /* LOV settings descriptor (should only contain static info) */
1802 __u32 ld_tgt_count; /* how many OBD's */
1803 __u32 ld_active_tgt_count; /* how many active */
1804 __u32 ld_default_stripe_count; /* how many objects are used */
1805 __u32 ld_pattern; /* default PATTERN_RAID0 */
1806 __u64 ld_default_stripe_size; /* in bytes */
1807 __u64 ld_default_stripe_offset; /* in bytes */
1808 __u32 ld_padding_0; /* unused */
1809 __u32 ld_qos_maxage; /* in second */
1810 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1811 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1812 struct obd_uuid ld_uuid;
1815 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1817 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1822 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1827 LDLM_BL_CALLBACK = 104,
1828 LDLM_CP_CALLBACK = 105,
1829 LDLM_GL_CALLBACK = 106,
1832 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1834 #define RES_NAME_SIZE 4
1835 struct ldlm_res_id {
1836 __u64 name[RES_NAME_SIZE];
1839 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1855 #define LCK_MODE_NUM 8
1865 #define LDLM_MIN_TYPE LDLM_PLAIN
1867 struct ldlm_extent {
1873 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
1874 struct ldlm_extent *ex2)
1876 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
1879 struct ldlm_inodebits {
1886 __u64 blocking_export; /* not actually used over the wire */
1887 __u32 blocking_pid; /* not actually used over the wire */
1891 /* it's important that the fields of the ldlm_extent structure match
1892 * the first fields of the ldlm_flock structure because there is only
1893 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1894 * this ever changes we will need to swab the union differently based
1895 * on the resource type. */
1898 struct ldlm_extent l_extent;
1899 struct ldlm_flock l_flock;
1900 struct ldlm_inodebits l_inodebits;
1901 } ldlm_policy_data_t;
1903 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1905 struct ldlm_intent {
1909 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1911 struct ldlm_resource_desc {
1912 ldlm_type_t lr_type;
1913 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1914 struct ldlm_res_id lr_name;
1917 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1919 struct ldlm_lock_desc {
1920 struct ldlm_resource_desc l_resource;
1921 ldlm_mode_t l_req_mode;
1922 ldlm_mode_t l_granted_mode;
1923 ldlm_policy_data_t l_policy_data;
1926 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1928 #define LDLM_LOCKREQ_HANDLES 2
1929 #define LDLM_ENQUEUE_CANCEL_OFF 1
1931 struct ldlm_request {
1934 struct ldlm_lock_desc lock_desc;
1935 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
1938 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1940 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
1941 * Otherwise, 2 are available. */
1942 #define ldlm_request_bufsize(count,type) \
1944 int _avail = LDLM_LOCKREQ_HANDLES; \
1945 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
1946 sizeof(struct ldlm_request) + \
1947 (count > _avail ? count - _avail : 0) * \
1948 sizeof(struct lustre_handle); \
1953 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1954 struct ldlm_lock_desc lock_desc;
1955 struct lustre_handle lock_handle;
1956 __u64 lock_policy_res1;
1957 __u64 lock_policy_res2;
1960 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1963 * Opcodes for mountconf (mgs and mgc)
1968 MGS_EXCEPTION, /* node died, etc. */
1969 MGS_TARGET_REG, /* whenever target starts up */
1974 #define MGS_FIRST_OPC MGS_CONNECT
1976 #define MGS_PARAM_MAXLEN 1024
1977 #define KEY_SET_INFO "set_info"
1979 struct mgs_send_param {
1980 char mgs_param[MGS_PARAM_MAXLEN];
1983 /* We pass this info to the MGS so it can write config logs */
1984 #define MTI_NAME_MAXLEN 64
1985 #define MTI_PARAM_MAXLEN 4096
1986 #define MTI_NIDS_MAX 32
1987 struct mgs_target_info {
1988 __u32 mti_lustre_ver;
1989 __u32 mti_stripe_index;
1990 __u32 mti_config_ver;
1992 __u32 mti_nid_count;
1993 __u32 padding; /* 64 bit align */
1994 char mti_fsname[MTI_NAME_MAXLEN];
1995 char mti_svname[MTI_NAME_MAXLEN];
1996 char mti_uuid[sizeof(struct obd_uuid)];
1997 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
1998 char mti_params[MTI_PARAM_MAXLEN];
2001 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
2003 /* Config marker flags (in config log) */
2004 #define CM_START 0x01
2006 #define CM_SKIP 0x04
2007 #define CM_UPGRADE146 0x08
2008 #define CM_EXCLUDE 0x10
2009 #define CM_START_SKIP (CM_START | CM_SKIP)
2012 __u32 cm_step; /* aka config version */
2014 __u32 cm_vers; /* lustre release version number */
2015 __u32 padding; /* 64 bit align */
2016 __u64 cm_createtime; /*when this record was first created */
2017 __u64 cm_canceltime; /*when this record is no longer valid*/
2018 char cm_tgtname[MTI_NAME_MAXLEN];
2019 char cm_comment[MTI_NAME_MAXLEN];
2022 extern void lustre_swab_cfg_marker(struct cfg_marker *marker,
2023 int swab, int size);
2027 * Opcodes for multiple servers.
2036 #define OBD_FIRST_OPC OBD_PING
2038 /* catalog of log objects */
2040 /** Identifier for a single log object */
2045 } __attribute__((packed));
2047 /** Records written to the CATALOGS list */
2048 #define CATLIST "CATALOGS"
2050 struct llog_logid lci_logid;
2054 } __attribute__((packed));
2056 /** join file lov mds md*/
2057 struct lov_mds_md_join {
2058 struct lov_mds_md lmmj_md;
2059 /*join private info*/
2060 struct llog_logid lmmj_array_id; /*array object id*/
2061 __u32 lmmj_extent_count; /*array extent count*/
2064 /* Log data record types - there is no specific reason that these need to
2065 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2067 #define LLOG_OP_MAGIC 0x10600000
2068 #define LLOG_OP_MASK 0xfff00000
2071 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2072 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2073 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
2074 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
2075 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
2076 MDS_SETATTR64_REC= LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) | REINT_SETATTR,
2077 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2078 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
2079 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2080 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
2081 /** changelog record type */
2082 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2083 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2084 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2088 * for now, continue to support old pad records which have 0 for their
2089 * type but still need to be swabbed for their length
2091 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2092 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
2093 __swab32(LLOG_OP_MAGIC) || \
2094 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
2096 /** Log record header - stored in little endian order.
2097 * Each record must start with this struct, end with a llog_rec_tail,
2098 * and be a multiple of 256 bits in size.
2100 struct llog_rec_hdr {
2107 struct llog_rec_tail {
2112 struct llog_logid_rec {
2113 struct llog_rec_hdr lid_hdr;
2114 struct llog_logid lid_id;
2120 struct llog_rec_tail lid_tail;
2121 } __attribute__((packed));
2123 /** MDS extent description
2124 * It is for joined file extent info, each extent info for joined file
2125 * just like (start, end, lmm).
2127 struct mds_extent_desc {
2128 __u64 med_start; /* extent start */
2129 __u64 med_len; /* extent length */
2130 struct lov_mds_md med_lmm; /* extent's lmm */
2133 /** Joined file array extent log record*/
2134 struct llog_array_rec {
2135 struct llog_rec_hdr lmr_hdr;
2136 struct mds_extent_desc lmr_med;
2137 struct llog_rec_tail lmr_tail;
2140 struct llog_create_rec {
2141 struct llog_rec_hdr lcr_hdr;
2142 struct ll_fid lcr_fid;
2146 struct llog_rec_tail lcr_tail;
2147 } __attribute__((packed));
2149 struct llog_orphan_rec {
2150 struct llog_rec_hdr lor_hdr;
2154 struct llog_rec_tail lor_tail;
2155 } __attribute__((packed));
2157 struct llog_unlink_rec {
2158 struct llog_rec_hdr lur_hdr;
2162 struct llog_rec_tail lur_tail;
2163 } __attribute__((packed));
2165 struct llog_setattr_rec {
2166 struct llog_rec_hdr lsr_hdr;
2172 struct llog_rec_tail lsr_tail;
2173 } __attribute__((packed));
2175 struct llog_setattr64_rec {
2176 struct llog_rec_hdr lsr_hdr;
2184 struct llog_rec_tail lsr_tail;
2185 } __attribute__((packed));
2187 struct llog_size_change_rec {
2188 struct llog_rec_hdr lsc_hdr;
2189 struct ll_fid lsc_fid;
2192 struct llog_rec_tail lsc_tail;
2193 } __attribute__((packed));
2195 #define CHANGELOG_MAGIC 0xca103000
2196 /** Changelog record types
2197 * When adding record types, update mdd_lproc.c's changelog_str
2199 enum changelog_rec_type {
2201 CL_CREATE = 1, /* namespace */
2202 CL_MKDIR = 2, /* namespace */
2203 CL_HARDLINK = 3, /* namespace */
2204 CL_SOFTLINK = 4, /* namespace */
2205 CL_MKNOD = 5, /* namespace */
2206 CL_UNLINK = 6, /* namespace */
2207 CL_RMDIR = 7, /* namespace */
2208 CL_RENAME = 8, /* namespace */
2209 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
2210 CL_OPEN = 10, /* not currently used */
2211 CL_CLOSE = 11, /* may be written to log only with mtime change */
2219 /** \a changelog_rec_type's that can't be masked */
2220 #define CL_MINMASK (1 << CL_MARK)
2221 /** bits covering all \a changelog_rec_type's */
2222 #define CL_ALLMASK 0XFFFF
2223 /** default \a changelog_rec_type mask */
2224 #define CL_DEFMASK CL_ALLMASK
2226 /* per-record flags */
2227 #define CLF_VERSION 0x1000
2228 #define CLF_FLAGMASK 0x0FFF
2229 #define CLF_HSM 0x0001
2231 /** changelog record */
2232 struct llog_changelog_rec {
2233 struct llog_rec_hdr cr_hdr;
2234 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
2236 __u32 cr_type; /**< \a changelog_rec_type */
2238 __u64 cr_prev; /**< last index for this target fid */
2241 struct lu_fid cr_tfid; /**< target fid */
2242 __u32 cr_markerflags; /**< CL_MARK flags */
2244 struct lu_fid cr_pfid; /**< parent fid */
2246 char cr_name[0]; /**< last element */
2247 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
2249 } __attribute__((packed));
2254 } __attribute__((packed));
2256 struct llog_gen_rec {
2257 struct llog_rec_hdr lgr_hdr;
2258 struct llog_gen lgr_gen;
2259 struct llog_rec_tail lgr_tail;
2261 /* On-disk header structure of each log object, stored in little endian order */
2262 #define LLOG_CHUNK_SIZE 8192
2263 #define LLOG_HEADER_SIZE (96)
2264 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2266 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2268 /* flags for the logs */
2269 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
2270 #define LLOG_F_IS_CAT 0x2
2271 #define LLOG_F_IS_PLAIN 0x4
2273 struct llog_log_hdr {
2274 struct llog_rec_hdr llh_hdr;
2275 __u64 llh_timestamp;
2277 __u32 llh_bitmap_offset;
2281 /* for a catalog the first plain slot is next to it */
2282 struct obd_uuid llh_tgtuuid;
2283 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2284 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2285 struct llog_rec_tail llh_tail;
2286 } __attribute__((packed));
2288 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
2289 llh->llh_bitmap_offset - \
2290 sizeof(llh->llh_tail)) * 8)
2292 /** log cookies are used to reference a specific log file and a record therein */
2293 struct llog_cookie {
2294 struct llog_logid lgc_lgl;
2298 } __attribute__((packed));
2300 /** llog protocol */
2301 enum llogd_rpc_ops {
2302 LLOG_ORIGIN_HANDLE_CREATE = 501,
2303 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2304 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2305 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2306 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2307 LLOG_ORIGIN_CONNECT = 506,
2308 LLOG_CATINFO = 507, /* for lfs catinfo */
2309 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2310 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2312 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2316 struct llog_logid lgd_logid;
2318 __u32 lgd_llh_flags;
2320 __u32 lgd_saved_index;
2322 __u64 lgd_cur_offset;
2323 } __attribute__((packed));
2325 struct llogd_conn_body {
2326 struct llog_gen lgdc_gen;
2327 struct llog_logid lgdc_logid;
2328 __u32 lgdc_ctxt_idx;
2329 } __attribute__((packed));
2331 struct lov_user_ost_data_join { /* per-stripe data structure */
2332 __u64 l_extent_start; /* extent start*/
2333 __u64 l_extent_end; /* extent end*/
2334 __u64 l_object_id; /* OST object ID */
2335 __u64 l_object_gr; /* OST object group (creating MDS number) */
2336 __u32 l_ost_gen; /* generation of this OST index */
2337 __u32 l_ost_idx; /* OST index in LOV */
2338 } __attribute__((packed));
2340 struct lov_user_md_join { /* LOV EA user data (host-endian) */
2341 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
2342 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
2343 __u64 lmm_object_id; /* LOV object ID */
2344 __u64 lmm_object_gr; /* LOV object group */
2345 __u32 lmm_stripe_size; /* size of stripe in bytes */
2346 __u32 lmm_stripe_count; /* num stripes in use for this object */
2347 __u32 lmm_extent_count; /* extent count of lmm*/
2348 __u64 lmm_tree_id; /* mds tree object id */
2349 __u64 lmm_tree_gen; /* mds tree object gen */
2350 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
2351 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
2352 } __attribute__((packed));
2354 /* Note: 64-bit types are 64-bit aligned in structure */
2356 obd_valid o_valid; /* hot fields in this obdo */
2360 obd_size o_size; /* o_size-o_blocks == ost_lvb */
2364 obd_blocks o_blocks; /* brw: cli sent cached bytes */
2367 /* 32-bit fields start here: keep an even number of them via padding */
2368 obd_blksize o_blksize; /* optimal IO blocksize */
2369 obd_mode o_mode; /* brw: cli sent cache remain */
2373 obd_count o_nlink; /* brw: checksum */
2374 obd_count o_generation;
2375 obd_count o_misc; /* brw: o_dropped */
2376 __u32 o_easize; /* epoch in ost writes */
2378 __u32 o_stripe_idx; /* holds stripe idx */
2380 struct lustre_handle o_handle; /* brw: lock handle to prolong locks */
2381 struct llog_cookie o_lcookie; /* destroy: unlink cookie from MDS */
2390 #define o_dirty o_blocks
2391 #define o_undirty o_mode
2392 #define o_dropped o_misc
2393 #define o_cksum o_nlink
2395 extern void lustre_swab_obdo (struct obdo *o);
2397 /* request structure for OST's */
2402 /* Key for FIEMAP to be used in get_info calls */
2403 struct ll_fiemap_info_key {
2406 struct ll_user_fiemap fiemap;
2409 extern void lustre_swab_ost_body (struct ost_body *b);
2410 extern void lustre_swab_ost_last_id(obd_id *id);
2411 extern void lustre_swab_fiemap(struct ll_user_fiemap *fiemap);
2413 extern void lustre_swab_lov_user_md_v1(struct lov_user_md_v1 *lum);
2414 extern void lustre_swab_lov_user_md_v3(struct lov_user_md_v3 *lum);
2415 extern void lustre_swab_lov_user_md_objects(struct lov_user_ost_data *lod,
2417 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
2420 extern void lustre_swab_llogd_body (struct llogd_body *d);
2421 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
2422 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
2423 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
2424 struct llog_rec_tail *tail);
2427 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
2429 /* this will be used when OBD_CONNECT_CHANGE_QS is set */
2432 * ID appiles to (uid, gid)
2436 * LQUOTA_FLAGS_* affect the responding bits
2440 * acquire/release count (bytes for block quota)
2444 * when a master returns the reply to a slave, it will
2445 * contain the current corresponding qunit size
2451 #define QDATA_IS_GRP(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_GRP)
2452 #define QDATA_IS_BLK(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_BLK)
2453 #define QDATA_IS_ADJBLK(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_ADJBLK)
2454 #define QDATA_IS_ADJINO(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_ADJINO)
2455 #define QDATA_IS_CHANGE_QS(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_CHG_QS)
2457 #define QDATA_SET_GRP(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_GRP)
2458 #define QDATA_SET_BLK(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_BLK)
2459 #define QDATA_SET_ADJBLK(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_ADJBLK)
2460 #define QDATA_SET_ADJINO(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_ADJINO)
2461 #define QDATA_SET_CHANGE_QS(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_CHG_QS)
2463 #define QDATA_CLR_GRP(qdata) ((qdata)->qd_flags &= ~LQUOTA_FLAGS_GRP)
2464 #define QDATA_CLR_CHANGE_QS(qdata) ((qdata)->qd_flags &= ~LQUOTA_FLAGS_CHG_QS)
2466 extern void lustre_swab_qdata(struct qunit_data *d);
2467 extern int quota_get_qdata(void*req, struct qunit_data *qdata,
2468 int is_req, int is_exp);
2469 extern int quota_copy_qdata(void *request, struct qunit_data *qdata,
2470 int is_req, int is_exp);
2477 #define QUOTA_FIRST_OPC QUOTA_DQACQ
2479 #define JOIN_FILE_ALIGN 4096
2481 #define QUOTA_REQUEST 1
2482 #define QUOTA_REPLY 0
2483 #define QUOTA_EXPORT 1
2484 #define QUOTA_IMPORT 0
2486 /* quota check function */
2487 #define QUOTA_RET_OK 0 /**< return successfully */
2488 #define QUOTA_RET_NOQUOTA 1 /**< not support quota */
2489 #define QUOTA_RET_NOLIMIT 2 /**< quota limit isn't set */
2490 #define QUOTA_RET_ACQUOTA 4 /**< need to acquire extra quota */
2491 #define QUOTA_RET_INC_PENDING 8 /**< pending value is increased */
2493 /* security opcodes */
2496 SEC_CTX_INIT_CONT = 802,
2499 SEC_FIRST_OPC = SEC_CTX_INIT
2503 * capa related definitions
2505 #define CAPA_HMAC_MAX_LEN 64
2506 #define CAPA_HMAC_KEY_MAX_LEN 56
2508 /* NB take care when changing the sequence of elements this struct,
2509 * because the offset info is used in find_capa() */
2510 struct lustre_capa {
2511 struct lu_fid lc_fid; /** fid */
2512 __u64 lc_opc; /** operations allowed */
2513 __u64 lc_uid; /** file owner */
2514 __u64 lc_gid; /** file group */
2515 __u32 lc_flags; /** HMAC algorithm & flags */
2516 __u32 lc_keyid; /** key# used for the capability */
2517 __u32 lc_timeout; /** capa timeout value (sec) */
2518 __u32 lc_expiry; /** expiry time (sec) */
2519 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
2520 } __attribute__((packed));
2522 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
2524 /** lustre_capa::lc_opc */
2526 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
2527 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
2528 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
2529 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
2530 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
2531 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
2532 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
2533 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
2534 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
2535 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
2536 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
2539 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2540 #define CAPA_OPC_MDS_ONLY \
2541 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2542 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2543 #define CAPA_OPC_OSS_ONLY \
2544 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
2545 CAPA_OPC_OSS_DESTROY)
2546 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2547 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2549 /* MDS capability covers object capability for operations of body r/w
2550 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
2551 * while OSS capability only covers object capability for operations of
2552 * oss data(file content) r/w/truncate.
2554 static inline int capa_for_mds(struct lustre_capa *c)
2556 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2559 static inline int capa_for_oss(struct lustre_capa *c)
2561 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2564 /* lustre_capa::lc_hmac_alg */
2566 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
2570 #define CAPA_FL_MASK 0x00ffffff
2571 #define CAPA_HMAC_ALG_MASK 0xff000000
2573 struct lustre_capa_key {
2574 __u64 lk_mdsid; /**< mds# */
2575 __u32 lk_keyid; /**< key# */
2577 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
2578 } __attribute__((packed));
2580 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2582 /** The link ea holds 1 \a link_ea_entry for each hardlink */
2583 #define LINK_EA_MAGIC 0x01EA0000
2584 struct link_ea_header {
2587 __u64 leh_len; /* total size */
2593 /** Hardlink data is name and parent fid.
2594 * Stored in this crazy struct for maximum packing and endian-neutrality
2596 struct link_ea_entry {
2597 /** __u16 stored big-endian, unaligned */
2599 struct lu_fid_pack lee_parent_fid; /**< variable length */
2600 /** logically after lee_parent_fid; don't use directly */