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 * ALL structs passing over the wire should be declared here. Structs
46 * that are used in interfaces with userspace should go in lustre_user.h.
48 * All structs being declared here should be built from simple fixed-size
49 * types (__u8, __u16, __u32, __u64) or be built from other types or
50 * structs also declared in this file. Similarly, all flags and magic
51 * values in those structs should also be declared here. This ensures
52 * that the Lustre wire protocol is not influenced by external dependencies.
54 * The only other acceptable items in this file are VERY SIMPLE accessor
55 * functions to avoid callers grubbing inside the structures, and the
56 * prototypes of the swabber functions for each struct. Nothing that
57 * depends on external functions or definitions should be in here.
59 * Structs must be properly aligned to put 64-bit values on an 8-byte
60 * boundary. Any structs being added here must also be added to
61 * utils/wirecheck.c and "make newwiretest" run to regenerate the
62 * utils/wiretest.c sources. This allows us to verify that wire structs
63 * have the proper alignment/size on all architectures.
65 * DO NOT CHANGE any of the structs, flags, values declared here and used
66 * in released Lustre versions. Some structs may have padding fields that
67 * can be used. Some structs might allow addition at the end (verify this
68 * in the code to ensure that new/old clients that see this larger struct
69 * do not fail, otherwise you need to implement protocol compatibility).
71 * We assume all nodes are either little-endian or big-endian, and we
72 * always send messages in the sender's native format. The receiver
73 * detects the message format by checking the 'magic' field of the message
74 * (see lustre_msg_swabbed() below).
76 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
77 * implemented either here, inline (trivial implementations) or in
78 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
79 * endian, in-place in the message buffer.
81 * A swabber takes a single pointer argument. The caller must already have
82 * verified that the length of the message buffer >= sizeof (type).
84 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
85 * may be defined that swabs just the variable part, after the caller has
86 * verified that the message buffer is large enough.
91 #ifndef _LUSTRE_IDL_H_
92 #define _LUSTRE_IDL_H_
94 #include <libcfs/libcfs.h>
96 /* Defn's shared with user-space. */
97 #include <lustre/lustre_user.h>
98 #include <lustre/ll_fiemap.h>
103 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
104 * FOO_REPLY_PORTAL is for incoming replies on the FOO
105 * FOO_BULK_PORTAL is for incoming bulk on the FOO
108 #define CONNMGR_REQUEST_PORTAL 1
109 #define CONNMGR_REPLY_PORTAL 2
110 //#define OSC_REQUEST_PORTAL 3
111 #define OSC_REPLY_PORTAL 4
112 //#define OSC_BULK_PORTAL 5
113 #define OST_IO_PORTAL 6
114 #define OST_CREATE_PORTAL 7
115 #define OST_BULK_PORTAL 8
116 //#define MDC_REQUEST_PORTAL 9
117 #define MDC_REPLY_PORTAL 10
118 //#define MDC_BULK_PORTAL 11
119 #define MDS_REQUEST_PORTAL 12
120 //#define MDS_REPLY_PORTAL 13
121 #define MDS_BULK_PORTAL 14
122 #define LDLM_CB_REQUEST_PORTAL 15
123 #define LDLM_CB_REPLY_PORTAL 16
124 #define LDLM_CANCEL_REQUEST_PORTAL 17
125 #define LDLM_CANCEL_REPLY_PORTAL 18
126 //#define PTLBD_REQUEST_PORTAL 19
127 //#define PTLBD_REPLY_PORTAL 20
128 //#define PTLBD_BULK_PORTAL 21
129 #define MDS_SETATTR_PORTAL 22
130 #define MDS_READPAGE_PORTAL 23
131 #define MDS_MDS_PORTAL 24
133 #define MGC_REPLY_PORTAL 25
134 #define MGS_REQUEST_PORTAL 26
135 #define MGS_REPLY_PORTAL 27
136 #define OST_REQUEST_PORTAL 28
137 #define FLD_REQUEST_PORTAL 29
138 #define SEQ_METADATA_PORTAL 30
139 #define SEQ_DATA_PORTAL 31
140 #define SEQ_CONTROLLER_PORTAL 32
145 #define SVC_RUNNING 8
146 #define SVC_STOPPING 16
147 #define SVC_STOPPED 32
150 #define PTL_RPC_MSG_REQUEST 4711
151 #define PTL_RPC_MSG_ERR 4712
152 #define PTL_RPC_MSG_REPLY 4713
154 /* DON'T use swabbed values of MAGIC as magic! */
155 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
156 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
158 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
159 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
161 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
163 #define PTLRPC_MSG_VERSION 0x00000003
164 #define LUSTRE_VERSION_MASK 0xffff0000
165 #define LUSTRE_OBD_VERSION 0x00010000
166 #define LUSTRE_MDS_VERSION 0x00020000
167 #define LUSTRE_OST_VERSION 0x00030000
168 #define LUSTRE_DLM_VERSION 0x00040000
169 #define LUSTRE_LOG_VERSION 0x00050000
170 #define LUSTRE_MGS_VERSION 0x00060000
172 typedef __u32 mdsno_t;
173 typedef __u64 seqno_t;
176 * Describes a range of sequence, lsr_start is included but lsr_end is
178 * Same structure is used in fld module where lsr_mdt field holds mdt id
182 struct lu_seq_range {
190 * returns width of given range \a r
193 static inline __u64 range_space(const struct lu_seq_range *range)
195 return range->lsr_end - range->lsr_start;
199 * initialize range to zero
202 static inline void range_init(struct lu_seq_range *range)
204 range->lsr_start = range->lsr_end = range->lsr_mdt = 0;
208 * check if given seq id \a s is within given range \a r
211 static inline int range_within(const struct lu_seq_range *range,
214 return s >= range->lsr_start && s < range->lsr_end;
218 * allocate \a w units of sequence from range \a from.
220 static inline void range_alloc(struct lu_seq_range *to,
221 struct lu_seq_range *from,
224 to->lsr_start = from->lsr_start;
225 to->lsr_end = from->lsr_start + width;
226 from->lsr_start += width;
229 static inline int range_is_sane(const struct lu_seq_range *range)
231 return (range->lsr_end >= range->lsr_start);
234 static inline int range_is_zero(const struct lu_seq_range *range)
236 return (range->lsr_start == 0 && range->lsr_end == 0);
239 static inline int range_is_exhausted(const struct lu_seq_range *range)
242 return range_space(range) == 0;
245 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x):%x"
247 #define PRANGE(range) \
248 (range)->lsr_start, \
252 /** \defgroup lu_fid lu_fid
258 * Fid is a cluster-wide unique identifier of a file or an object
259 * (stripe). Fids are never reused. Fids are transmitted across network (in
260 * the sender byte-ordering), and stored on disk in a packed form (struct
261 * lu_fid_pack) in a big-endian order.
265 * fid sequence. Sequence is a unit of migration: all files (objects)
266 * with fids from a given sequence are stored on the same
269 * Lustre should support 2 ^ 64 objects, thus even if one
270 * sequence has one object we will never reach this value.
273 /** fid number within sequence. */
276 * fid version, used to distinguish different versions (in the sense
277 * of snapshots, etc.) of the same file system object. Not currently
284 * Following struct for MDT attributes, that will be kept inode's EA.
285 * Introduced in 2.0 release (please see b15993, for details)
287 struct lustre_mdt_attrs {
288 /** FID of this inode */
289 struct lu_fid lma_self_fid;
290 /** SOM state, mdt/ost type, others */
292 /** total sectors in objects */
293 __u64 lma_som_sectors;
301 /** initial fid id value */
302 LUSTRE_FID_INIT_OID = 1UL
305 /** returns fid object sequence */
306 static inline __u64 fid_seq(const struct lu_fid *fid)
311 /** returns fid object id */
312 static inline __u32 fid_oid(const struct lu_fid *fid)
317 /** returns fid object version */
318 static inline __u32 fid_ver(const struct lu_fid *fid)
323 static inline void fid_zero(struct lu_fid *fid)
325 memset(fid, 0, sizeof(*fid));
328 /* Normal FID sequence starts from this value, i.e. 1<<33 */
329 #define FID_SEQ_START 0x200000000ULL
331 /* IDIF sequence starts from this value, i.e. 1<<32 */
332 #define IDIF_SEQ_START 0x100000000ULL
335 * Check if a fid is igif or not.
336 * \param fid the fid to be tested.
337 * \return true if the fid is a igif; otherwise false.
339 static inline int fid_is_igif(const struct lu_fid *fid)
341 return fid_seq(fid) > 0 && fid_seq(fid) < IDIF_SEQ_START;
345 * Check if a fid is idif or not.
346 * \param fid the fid to be tested.
347 * \return true if the fid is a idif; otherwise false.
349 static inline int fid_is_idif(const struct lu_fid *fid)
351 return fid_seq(fid) >= IDIF_SEQ_START && fid_seq(fid) < FID_SEQ_START;
355 * Get inode number from a igif.
356 * \param fid a igif to get inode number from.
357 * \return inode number for the igif.
359 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
365 * Get inode generation from a igif.
366 * \param fid a igif to get inode generation from.
367 * \return inode generation for the igif.
369 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
374 #define DFID "["LPX64":0x%x:0x%x]"
375 #define SFID "0x%llx:0x%x:0x%x"
382 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
384 /* check that all fields are converted */
385 CLASSERT(sizeof *src ==
386 sizeof fid_seq(src) +
387 sizeof fid_oid(src) + sizeof fid_ver(src));
388 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
389 dst->f_seq = cpu_to_le64(fid_seq(src));
390 dst->f_oid = cpu_to_le32(fid_oid(src));
391 dst->f_ver = cpu_to_le32(fid_ver(src));
394 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
396 /* check that all fields are converted */
397 CLASSERT(sizeof *src ==
398 sizeof fid_seq(src) +
399 sizeof fid_oid(src) + sizeof fid_ver(src));
400 dst->f_seq = le64_to_cpu(fid_seq(src));
401 dst->f_oid = le32_to_cpu(fid_oid(src));
402 dst->f_ver = le32_to_cpu(fid_ver(src));
403 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
406 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
408 /* check that all fields are converted */
409 CLASSERT(sizeof *src ==
410 sizeof fid_seq(src) +
411 sizeof fid_oid(src) + sizeof fid_ver(src));
412 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
413 dst->f_seq = cpu_to_be64(fid_seq(src));
414 dst->f_oid = cpu_to_be32(fid_oid(src));
415 dst->f_ver = cpu_to_be32(fid_ver(src));
418 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
420 /* check that all fields are converted */
421 CLASSERT(sizeof *src ==
422 sizeof fid_seq(src) +
423 sizeof fid_oid(src) + sizeof fid_ver(src));
424 dst->f_seq = be64_to_cpu(fid_seq(src));
425 dst->f_oid = be32_to_cpu(fid_oid(src));
426 dst->f_ver = be32_to_cpu(fid_ver(src));
427 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
431 * Storage representation for fids.
433 * Variable size, first byte contains the length of the whole record.
437 char fp_area[sizeof(struct lu_fid)];
440 void fid_pack(struct lu_fid_pack *pack, const struct lu_fid *fid,
441 struct lu_fid *befider);
442 int fid_unpack(const struct lu_fid_pack *pack, struct lu_fid *fid);
444 static inline int fid_is_sane(const struct lu_fid *fid)
448 ((fid_seq(fid) >= FID_SEQ_START && fid_oid(fid) != 0
449 && fid_ver(fid) == 0) ||
453 static inline int fid_is_zero(const struct lu_fid *fid)
455 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
458 extern void lustre_swab_lu_fid(struct lu_fid *fid);
459 extern void lustre_swab_lu_seq_range(struct lu_seq_range *range);
461 static inline int lu_fid_eq(const struct lu_fid *f0,
462 const struct lu_fid *f1)
464 /* Check that there is no alignment padding. */
465 CLASSERT(sizeof *f0 ==
466 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
467 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
468 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
469 return memcmp(f0, f1, sizeof *f0) == 0;
472 #define __diff_normalize(val0, val1) \
474 typeof(val0) __val0 = (val0); \
475 typeof(val1) __val1 = (val1); \
477 (__val0 == __val1 ? 0 : __val0 > __val1 ? +1 : -1); \
480 static inline int lu_fid_cmp(const struct lu_fid *f0,
481 const struct lu_fid *f1)
484 __diff_normalize(fid_seq(f0), fid_seq(f1)) ?:
485 __diff_normalize(fid_oid(f0), fid_oid(f1)) ?:
486 __diff_normalize(fid_ver(f0), fid_ver(f1));
491 /** \defgroup lu_dir lu_dir
494 * Layout of readdir pages, as transmitted on wire.
497 struct lu_fid lde_fid;
506 __u64 ldp_hash_start;
510 struct lu_dirent ldp_entries[0];
513 enum lu_dirpage_flags {
517 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
519 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
522 return dp->ldp_entries;
525 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
527 struct lu_dirent *next;
529 if (le16_to_cpu(ent->lde_reclen) != 0)
530 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
537 static inline int lu_dirent_size(struct lu_dirent *ent)
539 if (le16_to_cpu(ent->lde_reclen) == 0) {
540 return (sizeof(*ent) +
541 le16_to_cpu(ent->lde_namelen) + 7) & ~7;
543 return le16_to_cpu(ent->lde_reclen);
546 #define DIR_END_OFF 0xfffffffffffffffeULL
550 struct lustre_handle {
553 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
555 static inline int lustre_handle_is_used(struct lustre_handle *lh)
557 return lh->cookie != 0ull;
560 static inline int lustre_handle_equal(struct lustre_handle *lh1,
561 struct lustre_handle *lh2)
563 return lh1->cookie == lh2->cookie;
566 static inline void lustre_handle_copy(struct lustre_handle *tgt,
567 struct lustre_handle *src)
569 tgt->cookie = src->cookie;
572 /* flags for lm_flags */
573 #define MSGHDR_AT_SUPPORT 0x1
575 #define lustre_msg lustre_msg_v2
576 /* we depend on this structure to be 8-byte aligned */
577 /* this type is only endian-adjusted in lustre_unpack_msg() */
578 struct lustre_msg_v2 {
590 /* without gss, ptlrpc_body is put at the first buffer. */
591 #define PTLRPC_NUM_VERSIONS 4
593 struct lustre_handle pb_handle;
600 __u64 pb_last_committed;
605 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
606 __u32 pb_service_time; /* for rep, actual service time */
609 /* VBR: pre-versions */
610 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
611 /* padding for future needs */
615 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
617 /* message body offset for lustre_msg_v2 */
618 /* ptlrpc body offset in all request/reply messages */
619 #define MSG_PTLRPC_BODY_OFF 0
621 /* normal request/reply message record offset */
622 #define REQ_REC_OFF 1
623 #define REPLY_REC_OFF 1
625 /* ldlm request message body offset */
626 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
627 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
629 /* ldlm intent lock message body offset */
630 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
631 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
633 /* ldlm reply message body offset */
634 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
635 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
637 /* Flags that are operation-specific go in the top 16 bits. */
638 #define MSG_OP_FLAG_MASK 0xffff0000
639 #define MSG_OP_FLAG_SHIFT 16
641 /* Flags that apply to all requests are in the bottom 16 bits */
642 #define MSG_GEN_FLAG_MASK 0x0000ffff
643 #define MSG_LAST_REPLAY 0x0001
644 #define MSG_RESENT 0x0002
645 #define MSG_REPLAY 0x0004
646 /* #define MSG_AT_SUPPORT 0x0008
647 * This was used in early prototypes of adaptive timeouts, and while there
648 * shouldn't be any users of that code there also isn't a need for using this
649 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
650 #define MSG_DELAY_REPLAY 0x0010
651 #define MSG_VERSION_REPLAY 0x0020
652 #define MSG_REQ_REPLAY_DONE 0x0040
653 #define MSG_LOCK_REPLAY_DONE 0x0080
656 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
659 #define MSG_CONNECT_RECOVERING 0x00000001
660 #define MSG_CONNECT_RECONNECT 0x00000002
661 #define MSG_CONNECT_REPLAYABLE 0x00000004
662 //#define MSG_CONNECT_PEER 0x8
663 #define MSG_CONNECT_LIBCLIENT 0x00000010
664 #define MSG_CONNECT_INITIAL 0x00000020
665 #define MSG_CONNECT_ASYNC 0x00000040
666 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
667 #define MSG_CONNECT_TRANSNO 0x00000100 /* report transno */
670 #define OBD_CONNECT_RDONLY 0x1ULL /*client allowed read-only access*/
671 #define OBD_CONNECT_INDEX 0x2ULL /*connect to specific LOV idx */
672 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
673 #define OBD_CONNECT_GRANT 0x8ULL /*OSC acquires grant at connect */
674 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for client */
675 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
676 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO request portal */
677 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
678 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attributes */
679 #define OBD_CONNECT_CROW 0x200ULL /*MDS+OST create objects on write*/
680 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
681 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends initial transno */
682 #define OBD_CONNECT_IBITS 0x1000ULL /*support for inodebits locks */
683 #define OBD_CONNECT_JOIN 0x2000ULL /*files can be concatenated */
684 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server supports GetAttr By Fid */
685 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open handle on special nodes*/
686 #define OBD_CONNECT_RMT_CLIENT 0x00010000ULL /*Remote client */
687 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x00020000ULL /*Remote client by force */
688 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
689 #define OBD_CONNECT_QUOTA64 0x80000ULL /*64bit qunit_data.qd_count */
690 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
691 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
692 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
693 #define OBD_CONNECT_SOM 0x00800000ULL /*Size on MDS */
694 #define OBD_CONNECT_AT 0x01000000ULL /*client uses adaptive timeouts */
695 #define OBD_CONNECT_LRU_RESIZE 0x02000000ULL /*LRU resize feature. */
696 #define OBD_CONNECT_MDS_MDS 0x04000000ULL /*MDS-MDS connection */
697 #define OBD_CONNECT_REAL 0x08000000ULL /*real connection */
698 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*shrink/enlarge qunit b=10600 */
699 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos */
700 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
701 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
702 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
703 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
704 * and lustre/utils/wirecheck.c */
706 #ifdef HAVE_LRU_RESIZE_SUPPORT
707 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
709 #define LRU_RESIZE_CONNECT_FLAG 0
712 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
713 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
714 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
715 OBD_CONNECT_NODEVOH |/* OBD_CONNECT_ATTRFID |*/\
716 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
717 OBD_CONNECT_RMT_CLIENT | \
718 OBD_CONNECT_RMT_CLIENT_FORCE | \
719 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | \
720 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
721 LRU_RESIZE_CONNECT_FLAG | \
723 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
724 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
725 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
726 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
727 OBD_CONNECT_CANCELSET | OBD_CONNECT_AT | \
728 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_CKSUM | \
729 OBD_CONNECT_CHANGE_QS | \
730 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_RMT_CLIENT | \
731 OBD_CONNECT_RMT_CLIENT_FORCE | \
733 #define ECHO_CONNECT_SUPPORTED (0)
734 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT)
736 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
737 ((patch)<<8) + (fix))
738 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
739 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
740 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
741 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
743 /* This structure is used for both request and reply.
745 * If we eventually have separate connect data for different types, which we
746 * almost certainly will, then perhaps we stick a union in here. */
747 struct obd_connect_data {
748 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
749 __u32 ocd_version; /* lustre release version number */
750 __u32 ocd_grant; /* initial cache grant amount (bytes) */
751 __u32 ocd_index; /* LOV index to connect to */
752 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
753 __u64 ocd_ibits_known; /* inode bits this client understands */
754 __u32 ocd_nllu; /* non-local-lustre-user */
755 __u32 ocd_nllg; /* non-local-lustre-group */
756 __u64 ocd_transno; /* first transno from client to be replayed */
757 __u32 ocd_group; /* MDS group on OST */
758 __u32 ocd_cksum_types; /* supported checksum algorithms */
759 __u64 padding1; /* also fix lustre_swab_connect */
760 __u64 padding2; /* also fix lustre_swab_connect */
763 extern void lustre_swab_connect(struct obd_connect_data *ocd);
766 * Supported checksum algorithms. Up to 32 checksum types are supported.
767 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
768 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
769 * algorithm and also the OBD_FL_CKSUM* flags.
772 OBD_CKSUM_CRC32 = 0x00000001,
773 OBD_CKSUM_ADLER = 0x00000002,
777 * OST requests: OBDO & OBD request records
782 OST_REPLY = 0, /* reply ? */
800 OST_QUOTA_ADJUST_QUNIT = 20,
803 #define OST_FIRST_OPC OST_REPLY
805 typedef __u64 obd_id;
806 typedef __u64 obd_gr;
807 typedef __u64 obd_time;
808 typedef __u64 obd_size;
809 typedef __u64 obd_off;
810 typedef __u64 obd_blocks;
811 typedef __u64 obd_valid;
812 typedef __u32 obd_blksize;
813 typedef __u32 obd_mode;
814 typedef __u32 obd_uid;
815 typedef __u32 obd_gid;
816 typedef __u32 obd_flag;
817 typedef __u32 obd_count;
819 #define OBD_FL_INLINEDATA (0x00000001)
820 #define OBD_FL_OBDMDEXISTS (0x00000002)
821 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
822 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
823 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
824 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
825 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
826 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
827 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
828 #define OBD_FL_CREATE_CROW (0x00000400) /* object should be create on write */
831 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
832 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
835 #define OBD_FL_TRUNCLOCK (0x00000800)
840 #define OBD_FL_CKSUM_CRC32 (0x00001000)
841 #define OBD_FL_CKSUM_ADLER (0x00002000)
842 #define OBD_FL_CKSUM_ALL (OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER)
844 #define LOV_MAGIC_V1 0x0BD10BD0
845 #define LOV_MAGIC LOV_MAGIC_V1
846 #define LOV_MAGIC_JOIN 0x0BD20BD0
847 #define LOV_MAGIC_V3 0x0BD30BD0
849 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
850 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
851 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
852 #define LOV_PATTERN_CMOBD 0x200
854 #define LOV_OBJECT_GROUP_DEFAULT ~0ULL
855 #define LOV_OBJECT_GROUP_CLEAR 0ULL
857 #define lov_ost_data lov_ost_data_v1
858 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
859 __u64 l_object_id; /* OST object ID */
860 __u64 l_object_gr; /* OST object group (creating MDS number) */
861 __u32 l_ost_gen; /* generation of this l_ost_idx */
862 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
865 #define lov_mds_md lov_mds_md_v1
866 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
867 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
868 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
869 __u64 lmm_object_id; /* LOV object ID */
870 __u64 lmm_object_gr; /* LOV object group */
871 __u32 lmm_stripe_size; /* size of stripe in bytes */
872 __u32 lmm_stripe_count; /* num stripes in use for this object */
873 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
876 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
878 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
879 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
881 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
882 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
883 #define XATTR_USER_PREFIX "user."
884 #define XATTR_TRUSTED_PREFIX "trusted."
885 #define XATTR_SECURITY_PREFIX "security."
886 #define XATTR_LUSTRE_PREFIX "lustre."
888 #define XATTR_NAME_LOV "trusted.lov"
889 #define XATTR_NAME_LMA "trusted.lma"
890 #define XATTR_NAME_LMV "trusted.lmv"
891 #define XATTR_NAME_LINK "trusted.link"
894 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
895 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
896 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
897 __u64 lmm_object_id; /* LOV object ID */
898 __u64 lmm_object_gr; /* LOV object group */
899 __u32 lmm_stripe_size; /* size of stripe in bytes */
900 __u32 lmm_stripe_count; /* num stripes in use for this object */
901 char lmm_pool_name[LOV_MAXPOOLNAME]; /* must be 32bit aligned */
902 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
906 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
907 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
908 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
909 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
910 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
911 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
912 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
913 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
914 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
915 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
916 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
917 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
918 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
919 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
920 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
921 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
922 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
923 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
924 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
925 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
926 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
927 /*#define OBD_MD_FLOSCOPQ (0x00400000ULL) osc opaque data, never used */
928 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
929 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
930 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
931 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
932 /* ->mds if epoch opens or closes */
933 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
934 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
935 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
936 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
937 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
939 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
940 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
941 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
943 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
944 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
945 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
946 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
947 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
948 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
949 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
950 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
951 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
953 #define OBD_FL_TRUNC (0x0000200000000000ULL) /* for filter_truncate */
955 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
956 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
957 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
958 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
960 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
961 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
962 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
963 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
964 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
966 /* don't forget obdo_fid which is way down at the bottom so it can
967 * come after the definition of llog_cookie */
980 __u32 os_state; /* positive error code on server */
992 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
993 #define OBD_STATFS_NODELAY 0x0001 /* requests should be send without delay
994 * and resends for avoid deadlocks */
996 #define OBD_STATFS_FROM_CACHE 0x0002 /* the statfs callback should not update
999 /* ost_body.data values for OST_BRW */
1001 #define OBD_BRW_READ 0x01
1002 #define OBD_BRW_WRITE 0x02
1003 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1004 #define OBD_BRW_SYNC 0x08
1005 #define OBD_BRW_CHECK 0x10
1006 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1007 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1008 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1009 #define OBD_BRW_NOQUOTA 0x100
1010 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1012 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
1014 #define OST_MIN_PRECREATE 32
1015 #define OST_MAX_PRECREATE 20000
1024 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
1026 /* multiple of 8 bytes => can array */
1027 struct niobuf_remote {
1033 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
1035 /* lock value block communicated between the filter and llite */
1037 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1038 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1039 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1040 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1041 #define OST_LVB_IS_ERR(blocks) \
1042 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1043 #define OST_LVB_SET_ERR(blocks, rc) \
1044 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1045 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1055 extern void lustre_swab_ost_lvb(struct ost_lvb *);
1064 MDS_GETATTR_NAME = 34,
1069 MDS_DISCONNECT = 39,
1075 MDS_DONE_WRITING = 45,
1077 MDS_QUOTACHECK = 47,
1080 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1086 #define MDS_FIRST_OPC MDS_GETATTR
1103 } mds_reint_t, mdt_reint_t;
1105 extern void lustre_swab_generic_32s (__u32 *val);
1107 /* the disposition of the intent outlines what was executed */
1108 #define DISP_IT_EXECD 0x00000001
1109 #define DISP_LOOKUP_EXECD 0x00000002
1110 #define DISP_LOOKUP_NEG 0x00000004
1111 #define DISP_LOOKUP_POS 0x00000008
1112 #define DISP_OPEN_CREATE 0x00000010
1113 #define DISP_OPEN_OPEN 0x00000020
1114 #define DISP_ENQ_COMPLETE 0x00400000
1115 #define DISP_ENQ_OPEN_REF 0x00800000
1116 #define DISP_ENQ_CREATE_REF 0x01000000
1117 #define DISP_OPEN_LOCK 0x02000000
1119 /* INODE LOCK PARTS */
1120 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1121 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1122 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1124 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
1125 #define MDS_INODELOCK_MAXSHIFT 2
1126 /* This FULL lock is useful to take on unlink sort of operations */
1127 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1130 __u64 id; /* holds object id */
1131 __u32 generation; /* holds object generation */
1133 __u32 f_type; /* holds object type or stripe idx when passing it to
1134 * OST for saving into EA. */
1137 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1139 #define MDS_STATUS_CONN 1
1140 #define MDS_STATUS_LOV 2
1142 struct mds_status_req {
1147 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
1149 /* mdt_thread_info.mti_flags. */
1151 /* The flag indicates Size-on-MDS attributes are changed. */
1152 MF_SOM_CHANGE = (1 << 0),
1153 /* Flags indicates an epoch opens or closes. */
1154 MF_EPOCH_OPEN = (1 << 1),
1155 MF_EPOCH_CLOSE = (1 << 2),
1156 MF_MDC_CANCEL_FID1 = (1 << 3),
1157 MF_MDC_CANCEL_FID2 = (1 << 4),
1158 MF_MDC_CANCEL_FID3 = (1 << 5),
1159 MF_MDC_CANCEL_FID4 = (1 << 6),
1162 #define MF_SOM_LOCAL_FLAGS (MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID2 | \
1163 MF_MDC_CANCEL_FID3 | MF_MDC_CANCEL_FID4)
1165 #define MDS_BFLAG_UNCOMMITTED_WRITES 0x1
1166 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
1168 /* these should be identical to their EXT3_*_FL counterparts, and are
1169 * redefined here only to avoid dragging in ext3_fs.h */
1170 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
1171 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
1172 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
1173 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
1174 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1177 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
1178 * and we need to decode these into local S_* flags in the inode. Otherwise
1179 * we pass flags straight through (see bug 9486). */
1180 static inline int ll_ext_to_inode_flags(int flags)
1182 return (flags & MDS_BFLAG_EXT_FLAGS) ?
1183 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
1184 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
1185 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
1186 #if defined(S_DIRSYNC)
1187 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1189 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1190 (flags & ~MDS_BFLAG_EXT_FLAGS);
1193 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1194 * and we pass these straight through. Otherwise we need to convert from
1195 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1196 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1198 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1199 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1200 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1201 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1202 #if defined(S_DIRSYNC)
1203 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1205 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1210 * while mds_body is to interact with 1.6, mdt_body is to interact with 2.0.
1211 * both of them should have the same fields layout, because at client side
1212 * one could be dynamically cast to the other.
1214 * mdt_body has large size than mds_body, with unused padding (48 bytes)
1215 * at the end. client always use size of mdt_body to prepare request/reply
1216 * buffers, and actual data could be interepeted as mdt_body or mds_body
1222 struct lustre_handle handle;
1224 __u64 size; /* Offset, in the case of MDS_READPAGE */
1228 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1237 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1239 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1245 __u32 max_cookiesize;
1246 __u32 padding_4; /* also fix lustre_swab_mds_body */
1249 extern void lustre_swab_mds_body (struct mds_body *b);
1254 struct lustre_handle handle;
1256 __u64 size; /* Offset, in the case of MDS_READPAGE */
1260 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1262 __u64 ino; /* for 1.6 compatibility */
1269 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1271 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1272 __u32 generation; /* for 1.6 compatibility */
1277 __u32 max_cookiesize;
1278 __u32 padding_4; /* also fix lustre_swab_mdt_body */
1287 extern void lustre_swab_mdt_body (struct mdt_body *b);
1290 struct lustre_handle handle;
1296 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1298 #define Q_QUOTACHECK 0x800100
1299 #define Q_INITQUOTA 0x800101 /* init slave limits */
1300 #define Q_GETOINFO 0x800102 /* get obd quota info */
1301 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1302 #define Q_FINVALIDATE 0x800104 /* invalidate operational quotas */
1304 #define Q_TYPEMATCH(id, type) \
1305 ((id) == (type) || (id) == UGQUOTA)
1307 #define Q_TYPESET(oqc, type) Q_TYPEMATCH((oqc)->qc_type, type)
1309 #define Q_GETOCMD(oqc) \
1310 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1312 #define QCTL_COPY(out, in) \
1314 Q_COPY(out, in, qc_cmd); \
1315 Q_COPY(out, in, qc_type); \
1316 Q_COPY(out, in, qc_id); \
1317 Q_COPY(out, in, qc_stat); \
1318 Q_COPY(out, in, qc_dqinfo); \
1319 Q_COPY(out, in, qc_dqblk); \
1322 struct obd_quotactl {
1327 struct obd_dqinfo qc_dqinfo;
1328 struct obd_dqblk qc_dqblk;
1331 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1333 struct quota_adjust_qunit {
1340 extern void lustre_swab_quota_adjust_qunit(struct quota_adjust_qunit *q);
1342 /* flags in qunit_data and quota_adjust_qunit will use macroes below */
1343 #define LQUOTA_FLAGS_GRP 1UL /* 0 is user, 1 is group */
1344 #define LQUOTA_FLAGS_BLK 2UL /* 0 is inode, 1 is block */
1345 #define LQUOTA_FLAGS_ADJBLK 4UL /* adjust the block qunit size */
1346 #define LQUOTA_FLAGS_ADJINO 8UL /* adjust the inode qunit size */
1347 #define LQUOTA_FLAGS_CHG_QS 16UL /* indicate whether it has capability of
1348 * OBD_CONNECT_CHANGE_QS */
1350 /* the status of lqsk_flags in struct lustre_qunit_size_key */
1351 #define LQUOTA_QUNIT_FLAGS (LQUOTA_FLAGS_GRP | LQUOTA_FLAGS_BLK)
1353 #define QAQ_IS_GRP(qaq) ((qaq)->qaq_flags & LQUOTA_FLAGS_GRP)
1354 #define QAQ_IS_ADJBLK(qaq) ((qaq)->qaq_flags & LQUOTA_FLAGS_ADJBLK)
1355 #define QAQ_IS_ADJINO(qaq) ((qaq)->qaq_flags & LQUOTA_FLAGS_ADJINO)
1357 #define QAQ_SET_GRP(qaq) ((qaq)->qaq_flags |= LQUOTA_FLAGS_GRP)
1358 #define QAQ_SET_ADJBLK(qaq) ((qaq)->qaq_flags |= LQUOTA_FLAGS_ADJBLK)
1359 #define QAQ_SET_ADJINO(qaq) ((qaq)->qaq_flags |= LQUOTA_FLAGS_ADJINO)
1361 /* inode access permission for remote user, the inode info are omitted,
1362 * for client knows them. */
1363 struct mds_remote_perm {
1368 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1371 /* permissions for md_perm.mp_perm */
1373 CFS_SETUID_PERM = 0x01,
1374 CFS_SETGID_PERM = 0x02,
1375 CFS_SETGRP_PERM = 0x04,
1376 CFS_RMTACL_PERM = 0x08,
1377 CFS_RMTOWN_PERM = 0x10
1380 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1382 struct mdt_remote_perm {
1387 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1390 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1392 struct mds_rec_setattr {
1399 struct ll_fid sa_fid;
1400 __u64 sa_valid; /* MDS_ATTR_* attributes */
1407 __u32 sa_attr_flags;
1408 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1411 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1413 struct mdt_rec_setattr {
1423 __u32 sa_padding_1_h;
1424 struct lu_fid sa_fid;
1433 __u32 sa_attr_flags;
1441 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1444 * Attribute flags used in mds_rec_setattr::sa_valid.
1445 * The kernel's #defines for ATTR_* should not be used over the network
1446 * since the client and MDS may run different kernels (see bug 13828)
1447 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1449 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1450 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1451 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1452 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1453 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1454 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1455 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1456 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1457 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1458 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1459 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1460 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1461 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1462 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1463 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1464 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1467 #define FMODE_READ 00000001
1468 #define FMODE_WRITE 00000002
1471 #define FMODE_EPOCH 01000000
1472 #define FMODE_EPOCHLCK 02000000
1473 #define FMODE_SOM 04000000
1474 #define FMODE_CLOSED 0
1476 #define MDS_OPEN_CREATED 00000010
1477 #define MDS_OPEN_CROSS 00000020
1479 #define MDS_FMODE_EXEC 00000004
1480 #define MDS_OPEN_CREAT 00000100
1481 #define MDS_OPEN_EXCL 00000200
1482 #define MDS_OPEN_TRUNC 00001000
1483 #define MDS_OPEN_APPEND 00002000
1484 #define MDS_OPEN_SYNC 00010000
1485 #define MDS_OPEN_DIRECTORY 00200000
1487 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1488 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1489 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1490 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1491 * with default ACL */
1492 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1493 * actually, this is for create, not
1494 * conflict with other open flags */
1495 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1496 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1497 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1499 /* permission for create non-directory file */
1500 #define MAY_CREATE (1 << 7)
1501 /* permission for create directory file */
1502 #define MAY_LINK (1 << 8)
1503 /* permission for delete from the directory */
1504 #define MAY_UNLINK (1 << 9)
1505 /* source's permission for rename */
1506 #define MAY_RENAME_SRC (1 << 10)
1507 /* target's permission for rename */
1508 #define MAY_RENAME_TAR (1 << 11)
1509 /* part (parent's) VTX permission check */
1510 #define MAY_VTX_PART (1 << 12)
1511 /* full VTX permission check */
1512 #define MAY_VTX_FULL (1 << 13)
1513 /* lfs rgetfacl permission check */
1514 #define MAY_RGETFACL (1 << 14)
1517 MDS_CHECK_SPLIT = 1 << 0,
1518 MDS_CROSS_REF = 1 << 1,
1519 MDS_VTX_BYPASS = 1 << 2,
1520 MDS_PERM_BYPASS = 1 << 3,
1522 MDS_QUOTA_IGNORE = 1 << 5
1525 struct mds_rec_create {
1530 __u32 cr_flags; /* for use with open */
1532 struct ll_fid cr_fid;
1533 struct ll_fid cr_replayfid;
1537 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1538 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1539 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1540 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1541 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1544 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1546 struct mdt_rec_create {
1554 __u32 cr_suppgid1_h;
1556 __u32 cr_suppgid2_h;
1557 struct lu_fid cr_fid1;
1558 struct lu_fid cr_fid2;
1559 struct lustre_handle cr_old_handle; /* handle in case of open replay */
1563 __u64 cr_padding_1; /* pad for 64 bits*/
1566 __u32 cr_flags; /* for use with open */
1572 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1574 struct mds_rec_join {
1575 struct ll_fid jr_fid;
1579 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1581 struct mdt_rec_join {
1582 struct lu_fid jr_fid;
1586 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1588 struct mds_rec_link {
1595 struct ll_fid lk_fid1;
1596 struct ll_fid lk_fid2;
1598 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1599 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1600 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1601 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1604 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1606 struct mdt_rec_link {
1614 __u32 lk_suppgid1_h;
1616 __u32 lk_suppgid2_h;
1617 struct lu_fid lk_fid1;
1618 struct lu_fid lk_fid2;
1632 struct mds_rec_unlink {
1639 struct ll_fid ul_fid1;
1640 struct ll_fid ul_fid2;
1642 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1643 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1644 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1645 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1648 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1650 struct mdt_rec_unlink {
1658 __u32 ul_suppgid1_h;
1660 __u32 ul_suppgid2_h;
1661 struct lu_fid ul_fid1;
1662 struct lu_fid ul_fid2;
1676 struct mds_rec_rename {
1683 struct ll_fid rn_fid1;
1684 struct ll_fid rn_fid2;
1686 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1687 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1688 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1689 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1692 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1694 struct mdt_rec_rename {
1702 __u32 rn_suppgid1_h;
1704 __u32 rn_suppgid2_h;
1705 struct lu_fid rn_fid1;
1706 struct lu_fid rn_fid2;
1712 __u32 rn_bias; /* some operation flags */
1713 __u32 rn_mode; /* cross-ref rename has mode */
1720 struct mdt_rec_setxattr {
1728 __u32 sx_suppgid1_h;
1730 __u32 sx_suppgid2_h;
1731 struct lu_fid sx_fid;
1732 __u64 sx_padding_1; /* These three members are lu_fid size */
1744 __u32 sx_padding_10;
1745 __u32 sx_padding_11;
1748 struct mdt_rec_reint {
1756 __u32 rr_suppgid1_h;
1758 __u32 rr_suppgid2_h;
1759 struct lu_fid rr_fid1;
1760 struct lu_fid rr_fid2;
1768 __u32 rr_padding_1; /* also fix lustre_swab_mdt_rec_reint */
1769 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
1770 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
1771 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1774 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
1777 __u32 ld_tgt_count; /* how many MDS's */
1778 __u32 ld_active_tgt_count; /* how many active */
1779 __u32 ld_default_stripe_count; /* how many objects are used */
1780 __u32 ld_pattern; /* default MEA_MAGIC_* */
1781 __u64 ld_default_hash_size;
1782 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
1783 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
1784 __u32 ld_qos_maxage; /* in second */
1785 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
1786 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
1787 struct obd_uuid ld_uuid;
1790 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1792 /* TODO: lmv_stripe_md should contain mds capabilities for all slave fids */
1793 struct lmv_stripe_md {
1798 char mea_pool_name[LOV_MAXPOOLNAME];
1799 struct lu_fid mea_ids[0];
1802 extern void lustre_swab_lmv_stripe_md(struct lmv_stripe_md *mea);
1804 /* lmv structures */
1805 #define MEA_MAGIC_LAST_CHAR 0xb2221ca1
1806 #define MEA_MAGIC_ALL_CHARS 0xb222a11c
1807 #define MEA_MAGIC_HASH_SEGMENT 0xb222a11b
1809 #define MAX_HASH_SIZE_32 0x7fffffffUL
1810 #define MAX_HASH_SIZE 0x7fffffffffffffffULL
1811 #define MAX_HASH_HIGHEST_BIT 0x1000000000000000ULL
1818 extern void lustre_swab_md_fld (struct md_fld *mf);
1823 FLD_FIRST_OPC = FLD_QUERY
1829 SEQ_FIRST_OPC = SEQ_QUERY
1833 SEQ_ALLOC_SUPER = 0,
1838 * LOV data structures
1841 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
1842 #define LOV_MIN_STRIPE_SIZE (1<<LOV_MIN_STRIPE_BITS)
1843 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1844 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
1846 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1847 /* The size of the buffer the lov/mdc reserves for the
1848 * array of UUIDs returned by the MDS. With the current
1849 * protocol, this will limit the max number of OSTs per LOV */
1851 #define LOV_DESC_MAGIC 0xB0CCDE5C
1853 /* LOV settings descriptor (should only contain static info) */
1855 __u32 ld_tgt_count; /* how many OBD's */
1856 __u32 ld_active_tgt_count; /* how many active */
1857 __u32 ld_default_stripe_count; /* how many objects are used */
1858 __u32 ld_pattern; /* default PATTERN_RAID0 */
1859 __u64 ld_default_stripe_size; /* in bytes */
1860 __u64 ld_default_stripe_offset; /* in bytes */
1861 __u32 ld_padding_0; /* unused */
1862 __u32 ld_qos_maxage; /* in second */
1863 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1864 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1865 struct obd_uuid ld_uuid;
1868 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1870 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1875 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1880 LDLM_BL_CALLBACK = 104,
1881 LDLM_CP_CALLBACK = 105,
1882 LDLM_GL_CALLBACK = 106,
1885 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1887 #define RES_NAME_SIZE 4
1888 struct ldlm_res_id {
1889 __u64 name[RES_NAME_SIZE];
1892 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1908 #define LCK_MODE_NUM 8
1918 #define LDLM_MIN_TYPE LDLM_PLAIN
1920 struct ldlm_extent {
1926 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
1927 struct ldlm_extent *ex2)
1929 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
1932 struct ldlm_inodebits {
1939 __u64 blocking_export; /* not actually used over the wire */
1940 __u32 blocking_pid; /* not actually used over the wire */
1944 /* it's important that the fields of the ldlm_extent structure match
1945 * the first fields of the ldlm_flock structure because there is only
1946 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1947 * this ever changes we will need to swab the union differently based
1948 * on the resource type. */
1951 struct ldlm_extent l_extent;
1952 struct ldlm_flock l_flock;
1953 struct ldlm_inodebits l_inodebits;
1954 } ldlm_policy_data_t;
1956 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1958 struct ldlm_intent {
1962 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1964 struct ldlm_resource_desc {
1965 ldlm_type_t lr_type;
1966 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1967 struct ldlm_res_id lr_name;
1970 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1972 struct ldlm_lock_desc {
1973 struct ldlm_resource_desc l_resource;
1974 ldlm_mode_t l_req_mode;
1975 ldlm_mode_t l_granted_mode;
1976 ldlm_policy_data_t l_policy_data;
1979 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1981 #define LDLM_LOCKREQ_HANDLES 2
1982 #define LDLM_ENQUEUE_CANCEL_OFF 1
1984 struct ldlm_request {
1987 struct ldlm_lock_desc lock_desc;
1988 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
1991 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1993 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
1994 * Otherwise, 2 are available. */
1995 #define ldlm_request_bufsize(count,type) \
1997 int _avail = LDLM_LOCKREQ_HANDLES; \
1998 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
1999 sizeof(struct ldlm_request) + \
2000 (count > _avail ? count - _avail : 0) * \
2001 sizeof(struct lustre_handle); \
2006 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2007 struct ldlm_lock_desc lock_desc;
2008 struct lustre_handle lock_handle;
2009 __u64 lock_policy_res1;
2010 __u64 lock_policy_res2;
2013 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
2016 * Opcodes for mountconf (mgs and mgc)
2021 MGS_EXCEPTION, /* node died, etc. */
2022 MGS_TARGET_REG, /* whenever target starts up */
2027 #define MGS_FIRST_OPC MGS_CONNECT
2029 #define MGS_PARAM_MAXLEN 1024
2030 #define KEY_SET_INFO "set_info"
2032 struct mgs_send_param {
2033 char mgs_param[MGS_PARAM_MAXLEN];
2036 /* We pass this info to the MGS so it can write config logs */
2037 #define MTI_NAME_MAXLEN 64
2038 #define MTI_PARAM_MAXLEN 4096
2039 #define MTI_NIDS_MAX 32
2040 struct mgs_target_info {
2041 __u32 mti_lustre_ver;
2042 __u32 mti_stripe_index;
2043 __u32 mti_config_ver;
2045 __u32 mti_nid_count;
2046 __u32 padding; /* 64 bit align */
2047 char mti_fsname[MTI_NAME_MAXLEN];
2048 char mti_svname[MTI_NAME_MAXLEN];
2049 char mti_uuid[sizeof(struct obd_uuid)];
2050 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
2051 char mti_params[MTI_PARAM_MAXLEN];
2054 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
2056 /* Config marker flags (in config log) */
2057 #define CM_START 0x01
2059 #define CM_SKIP 0x04
2060 #define CM_UPGRADE146 0x08
2061 #define CM_EXCLUDE 0x10
2062 #define CM_START_SKIP (CM_START | CM_SKIP)
2065 __u32 cm_step; /* aka config version */
2067 __u32 cm_vers; /* lustre release version number */
2068 __u32 padding; /* 64 bit align */
2069 __u64 cm_createtime; /*when this record was first created */
2070 __u64 cm_canceltime; /*when this record is no longer valid*/
2071 char cm_tgtname[MTI_NAME_MAXLEN];
2072 char cm_comment[MTI_NAME_MAXLEN];
2075 extern void lustre_swab_cfg_marker(struct cfg_marker *marker,
2076 int swab, int size);
2079 * Opcodes for multiple servers.
2088 #define OBD_FIRST_OPC OBD_PING
2090 /* catalog of log objects */
2092 /** Identifier for a single log object */
2097 } __attribute__((packed));
2099 /** Records written to the CATALOGS list */
2100 #define CATLIST "CATALOGS"
2102 struct llog_logid lci_logid;
2106 } __attribute__((packed));
2108 /** join file lov mds md*/
2109 struct lov_mds_md_join {
2110 struct lov_mds_md lmmj_md;
2111 /*join private info*/
2112 struct llog_logid lmmj_array_id; /*array object id*/
2113 __u32 lmmj_extent_count; /*array extent count*/
2116 /* Log data record types - there is no specific reason that these need to
2117 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2119 #define LLOG_OP_MAGIC 0x10600000
2120 #define LLOG_OP_MASK 0xfff00000
2123 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2124 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2125 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
2126 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
2127 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
2128 MDS_SETATTR64_REC= LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) | REINT_SETATTR,
2129 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2130 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
2131 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2132 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
2133 /** changelog record type */
2134 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2135 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2136 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2140 * for now, continue to support old pad records which have 0 for their
2141 * type but still need to be swabbed for their length
2143 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2144 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
2145 __swab32(LLOG_OP_MAGIC) || \
2146 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
2148 /** Log record header - stored in little endian order.
2149 * Each record must start with this struct, end with a llog_rec_tail,
2150 * and be a multiple of 256 bits in size.
2152 struct llog_rec_hdr {
2159 struct llog_rec_tail {
2164 struct llog_logid_rec {
2165 struct llog_rec_hdr lid_hdr;
2166 struct llog_logid lid_id;
2172 struct llog_rec_tail lid_tail;
2173 } __attribute__((packed));
2175 /** MDS extent description
2176 * It is for joined file extent info, each extent info for joined file
2177 * just like (start, end, lmm).
2179 struct mds_extent_desc {
2180 __u64 med_start; /* extent start */
2181 __u64 med_len; /* extent length */
2182 struct lov_mds_md med_lmm; /* extent's lmm */
2185 /** Joined file array extent log record*/
2186 struct llog_array_rec {
2187 struct llog_rec_hdr lmr_hdr;
2188 struct mds_extent_desc lmr_med;
2189 struct llog_rec_tail lmr_tail;
2192 struct llog_create_rec {
2193 struct llog_rec_hdr lcr_hdr;
2194 struct ll_fid lcr_fid;
2198 struct llog_rec_tail lcr_tail;
2199 } __attribute__((packed));
2201 struct llog_orphan_rec {
2202 struct llog_rec_hdr lor_hdr;
2206 struct llog_rec_tail lor_tail;
2207 } __attribute__((packed));
2209 struct llog_unlink_rec {
2210 struct llog_rec_hdr lur_hdr;
2214 struct llog_rec_tail lur_tail;
2215 } __attribute__((packed));
2217 struct llog_setattr_rec {
2218 struct llog_rec_hdr lsr_hdr;
2224 struct llog_rec_tail lsr_tail;
2225 } __attribute__((packed));
2227 struct llog_setattr64_rec {
2228 struct llog_rec_hdr lsr_hdr;
2236 struct llog_rec_tail lsr_tail;
2237 } __attribute__((packed));
2239 struct llog_size_change_rec {
2240 struct llog_rec_hdr lsc_hdr;
2241 struct ll_fid lsc_fid;
2244 struct llog_rec_tail lsc_tail;
2245 } __attribute__((packed));
2247 #define CHANGELOG_MAGIC 0xca103000
2248 /** Changelog record types
2249 * When adding record types, update mdd_lproc.c's changelog_str
2251 enum changelog_rec_type {
2253 CL_CREATE = 1, /* namespace */
2254 CL_MKDIR = 2, /* namespace */
2255 CL_HARDLINK = 3, /* namespace */
2256 CL_SOFTLINK = 4, /* namespace */
2257 CL_MKNOD = 5, /* namespace */
2258 CL_UNLINK = 6, /* namespace */
2259 CL_RMDIR = 7, /* namespace */
2260 CL_RENAME = 8, /* namespace */
2261 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
2262 CL_OPEN = 10, /* not currently used */
2263 CL_CLOSE = 11, /* may be written to log only with mtime change */
2271 /** \a changelog_rec_type's that can't be masked */
2272 #define CL_MINMASK (1 << CL_MARK)
2273 /** bits covering all \a changelog_rec_type's */
2274 #define CL_ALLMASK 0XFFFF
2275 /** default \a changelog_rec_type mask */
2276 #define CL_DEFMASK CL_ALLMASK
2278 /* per-record flags */
2279 #define CLF_VERSION 0x1000
2280 #define CLF_FLAGMASK 0x0FFF
2281 #define CLF_HSM 0x0001
2283 /** changelog record */
2284 struct llog_changelog_rec {
2285 struct llog_rec_hdr cr_hdr;
2286 __u16 cr_flags; /**< (flags&CLF_FLAGMASK)|CLF_VERSION */
2288 __u32 cr_type; /**< \a changelog_rec_type */
2290 __u64 cr_prev; /**< last index for this target fid */
2293 struct lu_fid cr_tfid; /**< target fid */
2294 __u32 cr_markerflags; /**< CL_MARK flags */
2296 struct lu_fid cr_pfid; /**< parent fid */
2298 char cr_name[0]; /**< last element */
2299 struct llog_rec_tail cr_tail; /**< for_sizezof_only */
2301 } __attribute__((packed));
2306 } __attribute__((packed));
2308 struct llog_gen_rec {
2309 struct llog_rec_hdr lgr_hdr;
2310 struct llog_gen lgr_gen;
2311 struct llog_rec_tail lgr_tail;
2313 /* On-disk header structure of each log object, stored in little endian order */
2314 #define LLOG_CHUNK_SIZE 8192
2315 #define LLOG_HEADER_SIZE (96)
2316 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2318 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2320 /* flags for the logs */
2321 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
2322 #define LLOG_F_IS_CAT 0x2
2323 #define LLOG_F_IS_PLAIN 0x4
2325 struct llog_log_hdr {
2326 struct llog_rec_hdr llh_hdr;
2327 __u64 llh_timestamp;
2329 __u32 llh_bitmap_offset;
2333 /* for a catalog the first plain slot is next to it */
2334 struct obd_uuid llh_tgtuuid;
2335 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2336 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2337 struct llog_rec_tail llh_tail;
2338 } __attribute__((packed));
2340 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
2341 llh->llh_bitmap_offset - \
2342 sizeof(llh->llh_tail)) * 8)
2344 /** log cookies are used to reference a specific log file and a record therein */
2345 struct llog_cookie {
2346 struct llog_logid lgc_lgl;
2350 } __attribute__((packed));
2352 /** llog protocol */
2353 enum llogd_rpc_ops {
2354 LLOG_ORIGIN_HANDLE_CREATE = 501,
2355 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2356 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2357 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2358 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2359 LLOG_ORIGIN_CONNECT = 506,
2360 LLOG_CATINFO = 507, /* for lfs catinfo */
2361 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2362 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2364 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2368 struct llog_logid lgd_logid;
2370 __u32 lgd_llh_flags;
2372 __u32 lgd_saved_index;
2374 __u64 lgd_cur_offset;
2375 } __attribute__((packed));
2377 struct llogd_conn_body {
2378 struct llog_gen lgdc_gen;
2379 struct llog_logid lgdc_logid;
2380 __u32 lgdc_ctxt_idx;
2381 } __attribute__((packed));
2383 struct lov_user_ost_data_join { /* per-stripe data structure */
2384 __u64 l_extent_start; /* extent start*/
2385 __u64 l_extent_end; /* extent end*/
2386 __u64 l_object_id; /* OST object ID */
2387 __u64 l_object_gr; /* OST object group (creating MDS number) */
2388 __u32 l_ost_gen; /* generation of this OST index */
2389 __u32 l_ost_idx; /* OST index in LOV */
2390 } __attribute__((packed));
2392 struct lov_user_md_join { /* LOV EA user data (host-endian) */
2393 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
2394 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
2395 __u64 lmm_object_id; /* LOV object ID */
2396 __u64 lmm_object_gr; /* LOV object group */
2397 __u32 lmm_stripe_size; /* size of stripe in bytes */
2398 __u32 lmm_stripe_count; /* num stripes in use for this object */
2399 __u32 lmm_extent_count; /* extent count of lmm*/
2400 __u64 lmm_tree_id; /* mds tree object id */
2401 __u64 lmm_tree_gen; /* mds tree object gen */
2402 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
2403 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
2404 } __attribute__((packed));
2406 /* Note: 64-bit types are 64-bit aligned in structure */
2408 obd_valid o_valid; /* hot fields in this obdo */
2412 obd_size o_size; /* o_size-o_blocks == ost_lvb */
2416 obd_blocks o_blocks; /* brw: cli sent cached bytes */
2419 /* 32-bit fields start here: keep an even number of them via padding */
2420 obd_blksize o_blksize; /* optimal IO blocksize */
2421 obd_mode o_mode; /* brw: cli sent cache remain */
2425 obd_count o_nlink; /* brw: checksum */
2426 obd_count o_generation;
2427 obd_count o_misc; /* brw: o_dropped */
2428 __u32 o_easize; /* epoch in ost writes */
2430 __u32 o_stripe_idx; /* holds stripe idx */
2432 struct lustre_handle o_handle; /* brw: lock handle to prolong locks */
2433 struct llog_cookie o_lcookie; /* destroy: unlink cookie from MDS */
2442 #define o_dirty o_blocks
2443 #define o_undirty o_mode
2444 #define o_dropped o_misc
2445 #define o_cksum o_nlink
2447 extern void lustre_swab_obdo (struct obdo *o);
2449 /* request structure for OST's */
2454 /* Key for FIEMAP to be used in get_info calls */
2455 struct ll_fiemap_info_key {
2458 struct ll_user_fiemap fiemap;
2461 extern void lustre_swab_ost_body (struct ost_body *b);
2462 extern void lustre_swab_ost_last_id(obd_id *id);
2463 extern void lustre_swab_fiemap(struct ll_user_fiemap *fiemap);
2465 extern void lustre_swab_lov_user_md_v1(struct lov_user_md_v1 *lum);
2466 extern void lustre_swab_lov_user_md_v3(struct lov_user_md_v3 *lum);
2467 extern void lustre_swab_lov_user_md_objects(struct lov_user_ost_data *lod,
2469 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
2470 extern void lustre_swab_lov_mds_md(struct lov_mds_md *lmm);
2473 extern void lustre_swab_llogd_body (struct llogd_body *d);
2474 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
2475 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
2476 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
2477 struct llog_rec_tail *tail);
2480 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
2482 /* this will be used when OBD_CONNECT_CHANGE_QS is set */
2485 * ID appiles to (uid, gid)
2489 * LQUOTA_FLAGS_* affect the responding bits
2493 * acquire/release count (bytes for block quota)
2497 * when a master returns the reply to a slave, it will
2498 * contain the current corresponding qunit size
2504 #define QDATA_IS_GRP(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_GRP)
2505 #define QDATA_IS_BLK(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_BLK)
2506 #define QDATA_IS_ADJBLK(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_ADJBLK)
2507 #define QDATA_IS_ADJINO(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_ADJINO)
2508 #define QDATA_IS_CHANGE_QS(qdata) ((qdata)->qd_flags & LQUOTA_FLAGS_CHG_QS)
2510 #define QDATA_SET_GRP(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_GRP)
2511 #define QDATA_SET_BLK(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_BLK)
2512 #define QDATA_SET_ADJBLK(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_ADJBLK)
2513 #define QDATA_SET_ADJINO(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_ADJINO)
2514 #define QDATA_SET_CHANGE_QS(qdata) ((qdata)->qd_flags |= LQUOTA_FLAGS_CHG_QS)
2516 #define QDATA_CLR_GRP(qdata) ((qdata)->qd_flags &= ~LQUOTA_FLAGS_GRP)
2517 #define QDATA_CLR_CHANGE_QS(qdata) ((qdata)->qd_flags &= ~LQUOTA_FLAGS_CHG_QS)
2519 extern void lustre_swab_qdata(struct qunit_data *d);
2520 extern int quota_get_qdata(void*req, struct qunit_data *qdata,
2521 int is_req, int is_exp);
2522 extern int quota_copy_qdata(void *request, struct qunit_data *qdata,
2523 int is_req, int is_exp);
2530 #define QUOTA_FIRST_OPC QUOTA_DQACQ
2532 #define JOIN_FILE_ALIGN 4096
2534 #define QUOTA_REQUEST 1
2535 #define QUOTA_REPLY 0
2536 #define QUOTA_EXPORT 1
2537 #define QUOTA_IMPORT 0
2539 /* quota check function */
2540 #define QUOTA_RET_OK 0 /**< return successfully */
2541 #define QUOTA_RET_NOQUOTA 1 /**< not support quota */
2542 #define QUOTA_RET_NOLIMIT 2 /**< quota limit isn't set */
2543 #define QUOTA_RET_ACQUOTA 4 /**< need to acquire extra quota */
2546 /* security opcodes */
2549 SEC_CTX_INIT_CONT = 802,
2552 SEC_FIRST_OPC = SEC_CTX_INIT
2556 * capa related definitions
2558 #define CAPA_HMAC_MAX_LEN 64
2559 #define CAPA_HMAC_KEY_MAX_LEN 56
2561 /* NB take care when changing the sequence of elements this struct,
2562 * because the offset info is used in find_capa() */
2563 struct lustre_capa {
2564 struct lu_fid lc_fid; /** fid */
2565 __u64 lc_opc; /** operations allowed */
2566 __u64 lc_uid; /** file owner */
2567 __u64 lc_gid; /** file group */
2568 __u32 lc_flags; /** HMAC algorithm & flags */
2569 __u32 lc_keyid; /** key# used for the capability */
2570 __u32 lc_timeout; /** capa timeout value (sec) */
2571 __u32 lc_expiry; /** expiry time (sec) */
2572 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
2573 } __attribute__((packed));
2575 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
2577 /** lustre_capa::lc_opc */
2579 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
2580 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
2581 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
2582 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
2583 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
2584 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
2585 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
2586 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
2587 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
2588 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
2589 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
2592 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2593 #define CAPA_OPC_MDS_ONLY \
2594 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2595 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2596 #define CAPA_OPC_OSS_ONLY \
2597 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
2598 CAPA_OPC_OSS_DESTROY)
2599 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2600 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2602 /* MDS capability covers object capability for operations of body r/w
2603 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
2604 * while OSS capability only covers object capability for operations of
2605 * oss data(file content) r/w/truncate.
2607 static inline int capa_for_mds(struct lustre_capa *c)
2609 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2612 static inline int capa_for_oss(struct lustre_capa *c)
2614 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2617 /* lustre_capa::lc_hmac_alg */
2619 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
2623 #define CAPA_FL_MASK 0x00ffffff
2624 #define CAPA_HMAC_ALG_MASK 0xff000000
2626 struct lustre_capa_key {
2627 __u64 lk_mdsid; /**< mds# */
2628 __u32 lk_keyid; /**< key# */
2630 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
2631 } __attribute__((packed));
2633 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2635 /** The link ea holds 1 \a link_ea_entry for each hardlink */
2636 #define LINK_EA_MAGIC 0x01EA0000
2637 struct link_ea_header {
2640 __u64 leh_len; /* total size */
2646 /** Hardlink data is name and parent fid.
2647 * Stored in this crazy struct for maximum packing and endian-neutrality
2649 struct link_ea_entry {
2650 /** __u16 stored big-endian, unaligned */
2652 struct lu_fid_pack lee_parent_fid; /**< variable length */
2653 /** logically after lee_parent_fid; don't use directly */