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 [sun.com URL with a
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 * We assume all nodes are either little-endian or big-endian, and we
42 * always send messages in the sender's native format. The receiver
43 * detects the message format by checking the 'magic' field of the message
44 * (see lustre_msg_swabbed() below).
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 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
72 * implemented either here, inline (trivial implementations) or in
73 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
74 * endian, in-place in the message buffer.
76 * A swabber takes a single pointer argument. The caller must already have
77 * verified that the length of the message buffer >= sizeof (type).
79 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
80 * may be defined that swabs just the variable part, after the caller has
81 * verified that the message buffer is large enough.
84 #ifndef _LUSTRE_IDL_H_
85 #define _LUSTRE_IDL_H_
87 #include <libcfs/libcfs.h>
89 /* Defn's shared with user-space. */
90 #include <lustre/lustre_user.h>
95 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
96 * FOO_REPLY_PORTAL is for incoming replies on the FOO
97 * FOO_BULK_PORTAL is for incoming bulk on the FOO
100 #define CONNMGR_REQUEST_PORTAL 1
101 #define CONNMGR_REPLY_PORTAL 2
102 //#define OSC_REQUEST_PORTAL 3
103 #define OSC_REPLY_PORTAL 4
104 //#define OSC_BULK_PORTAL 5
105 #define OST_IO_PORTAL 6
106 #define OST_CREATE_PORTAL 7
107 #define OST_BULK_PORTAL 8
108 //#define MDC_REQUEST_PORTAL 9
109 #define MDC_REPLY_PORTAL 10
110 //#define MDC_BULK_PORTAL 11
111 #define MDS_REQUEST_PORTAL 12
112 //#define MDS_REPLY_PORTAL 13
113 #define MDS_BULK_PORTAL 14
114 #define LDLM_CB_REQUEST_PORTAL 15
115 #define LDLM_CB_REPLY_PORTAL 16
116 #define LDLM_CANCEL_REQUEST_PORTAL 17
117 #define LDLM_CANCEL_REPLY_PORTAL 18
118 //#define PTLBD_REQUEST_PORTAL 19
119 //#define PTLBD_REPLY_PORTAL 20
120 //#define PTLBD_BULK_PORTAL 21
121 #define MDS_SETATTR_PORTAL 22
122 #define MDS_READPAGE_PORTAL 23
123 #define MDS_MDS_PORTAL 24
125 #define MGC_REPLY_PORTAL 25
126 #define MGS_REQUEST_PORTAL 26
127 #define MGS_REPLY_PORTAL 27
128 #define OST_REQUEST_PORTAL 28
129 #define FLD_REQUEST_PORTAL 29
130 #define SEQ_METADATA_PORTAL 30
131 #define SEQ_DATA_PORTAL 31
132 #define SEQ_CONTROLLER_PORTAL 32
137 #define SVC_RUNNING 8
138 #define SVC_STOPPING 16
139 #define SVC_STOPPED 32
142 #define PTL_RPC_MSG_REQUEST 4711
143 #define PTL_RPC_MSG_ERR 4712
144 #define PTL_RPC_MSG_REPLY 4713
146 /* DON'T use swabbed values of MAGIC as magic! */
147 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
148 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
150 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
151 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
153 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
155 #define PTLRPC_MSG_VERSION 0x00000003
156 #define LUSTRE_VERSION_MASK 0xffff0000
157 #define LUSTRE_OBD_VERSION 0x00010000
158 #define LUSTRE_MDS_VERSION 0x00020000
159 #define LUSTRE_OST_VERSION 0x00030000
160 #define LUSTRE_DLM_VERSION 0x00040000
161 #define LUSTRE_LOG_VERSION 0x00050000
162 #define LUSTRE_MGS_VERSION 0x00060000
164 typedef __u64 mdsno_t;
165 typedef __u64 seqno_t;
172 static inline __u64 range_space(struct lu_range *r)
174 return r->lr_end - r->lr_start;
177 static inline void range_zero(struct lu_range *r)
179 r->lr_start = r->lr_end = 0;
182 static inline int range_within(struct lu_range *r,
185 return s >= r->lr_start && s < r->lr_end;
188 static inline void range_alloc(struct lu_range *r,
192 r->lr_start = s->lr_start;
193 r->lr_end = s->lr_start + w;
197 static inline int range_is_sane(struct lu_range *r)
199 return (r->lr_end >= r->lr_start);
202 static inline int range_is_zero(struct lu_range *r)
204 return (r->lr_start == 0 && r->lr_end == 0);
207 static inline int range_is_exhausted(struct lu_range *r)
209 return range_space(r) == 0;
212 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x]"
214 #define PRANGE(range) \
219 __u64 f_seq; /* holds fid sequence. Lustre should support 2 ^ 64
220 * objects, thus even if one sequence has one object we
221 * will never reach this value. */
222 __u32 f_oid; /* fid number within its sequence. */
223 __u32 f_ver; /* holds fid version. */
230 /* initial fid id value */
231 LUSTRE_FID_INIT_OID = 1UL
234 /* get object sequence */
235 static inline __u64 fid_seq(const struct lu_fid *fid)
241 static inline __u32 fid_oid(const struct lu_fid *fid)
246 /* get object version */
247 static inline __u32 fid_ver(const struct lu_fid *fid)
252 static inline void fid_zero(struct lu_fid *fid)
254 memset(fid, 0, sizeof(*fid));
257 /* Normal FID sequence starts from this value, i.e. 1<<33 */
258 #define FID_SEQ_START 0x200000000ULL
260 /* IDIF sequence starts from this value, i.e. 1<<32 */
261 #define IDIF_SEQ_START 0x100000000ULL
264 * Check if a fid is igif or not.
265 * \param fid the fid to be tested.
266 * \return true if the fid is a igif; otherwise false.
268 static inline int fid_is_igif(const struct lu_fid *fid)
270 return fid_seq(fid) > 0 && fid_seq(fid) < IDIF_SEQ_START;
274 * Check if a fid is idif or not.
275 * \param fid the fid to be tested.
276 * \return true if the fid is a idif; otherwise false.
278 static inline int fid_is_idif(const struct lu_fid *fid)
280 return fid_seq(fid) >= IDIF_SEQ_START && fid_seq(fid) < FID_SEQ_START;
284 * Get inode number from a igif.
285 * \param fid a igif to get inode number from.
286 * \return inode number for the igif.
288 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
294 * Get inode generation from a igif.
295 * \param fid a igif to get inode generation from.
296 * \return inode generation for the igif.
298 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
303 #define DFID "[0x%16.16"LPF64"x/0x%8.8x:0x%8.8x]"
310 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
312 /* check that all fields are converted */
313 CLASSERT(sizeof *src ==
314 sizeof fid_seq(src) +
315 sizeof fid_oid(src) + sizeof fid_ver(src));
316 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
317 dst->f_seq = cpu_to_le64(fid_seq(src));
318 dst->f_oid = cpu_to_le32(fid_oid(src));
319 dst->f_ver = cpu_to_le32(fid_ver(src));
322 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
324 /* check that all fields are converted */
325 CLASSERT(sizeof *src ==
326 sizeof fid_seq(src) +
327 sizeof fid_oid(src) + sizeof fid_ver(src));
328 dst->f_seq = le64_to_cpu(fid_seq(src));
329 dst->f_oid = le32_to_cpu(fid_oid(src));
330 dst->f_ver = le32_to_cpu(fid_ver(src));
331 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
334 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
336 /* check that all fields are converted */
337 CLASSERT(sizeof *src ==
338 sizeof fid_seq(src) +
339 sizeof fid_oid(src) + sizeof fid_ver(src));
340 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
341 dst->f_seq = cpu_to_be64(fid_seq(src));
342 dst->f_oid = cpu_to_be32(fid_oid(src));
343 dst->f_ver = cpu_to_be32(fid_ver(src));
346 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
348 /* check that all fields are converted */
349 CLASSERT(sizeof *src ==
350 sizeof fid_seq(src) +
351 sizeof fid_oid(src) + sizeof fid_ver(src));
352 dst->f_seq = be64_to_cpu(fid_seq(src));
353 dst->f_oid = be32_to_cpu(fid_oid(src));
354 dst->f_ver = be32_to_cpu(fid_ver(src));
355 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
360 * Storage representation for fids.
362 * Variable size, first byte contains the length of the whole record.
367 char fp_area[sizeof(struct lu_fid)];
370 void fid_pack(struct lu_fid_pack *pack, const struct lu_fid *fid,
371 struct lu_fid *befider);
372 int fid_unpack(const struct lu_fid_pack *pack, struct lu_fid *fid);
377 static inline int fid_is_sane(const struct lu_fid *fid)
381 ((fid_seq(fid) >= FID_SEQ_START && fid_oid(fid) != 0
382 && fid_ver(fid) == 0) ||
386 static inline int fid_is_zero(const struct lu_fid *fid)
388 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
391 extern void lustre_swab_lu_fid(struct lu_fid *fid);
392 extern void lustre_swab_lu_range(struct lu_range *range);
394 static inline int lu_fid_eq(const struct lu_fid *f0,
395 const struct lu_fid *f1)
397 /* Check that there is no alignment padding. */
398 CLASSERT(sizeof *f0 ==
399 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
400 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
401 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
402 return memcmp(f0, f1, sizeof *f0) == 0;
406 * Layout of readdir pages, as transmitted on wire.
409 struct lu_fid lde_fid;
418 __u64 ldp_hash_start;
422 struct lu_dirent ldp_entries[0];
425 enum lu_dirpage_flags {
429 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
431 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
434 return dp->ldp_entries;
437 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
439 struct lu_dirent *next;
441 if (le16_to_cpu(ent->lde_reclen) != 0)
442 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
449 static inline int lu_dirent_size(struct lu_dirent *ent)
451 if (le16_to_cpu(ent->lde_reclen) == 0) {
452 return (sizeof(*ent) +
453 le16_to_cpu(ent->lde_namelen) + 7) & ~7;
455 return le16_to_cpu(ent->lde_reclen);
458 #define DIR_END_OFF 0xfffffffffffffffeULL
460 struct lustre_handle {
463 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
465 static inline int lustre_handle_is_used(struct lustre_handle *lh)
467 return lh->cookie != 0ull;
470 static inline int lustre_handle_equal(struct lustre_handle *lh1,
471 struct lustre_handle *lh2)
473 return lh1->cookie == lh2->cookie;
476 static inline void lustre_handle_copy(struct lustre_handle *tgt,
477 struct lustre_handle *src)
479 tgt->cookie = src->cookie;
482 /* flags for lm_flags */
483 #define MSGHDR_AT_SUPPORT 0x1
485 #define lustre_msg lustre_msg_v2
486 /* we depend on this structure to be 8-byte aligned */
487 /* this type is only endian-adjusted in lustre_unpack_msg() */
488 struct lustre_msg_v2 {
500 /* without gss, ptlrpc_body is put at the first buffer. */
502 struct lustre_handle pb_handle;
509 __u64 pb_last_committed;
514 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
515 __u32 pb_service_time; /* for rep, actual service time */
520 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
522 /* message body offset for lustre_msg_v2 */
523 /* ptlrpc body offset in all request/reply messages */
524 #define MSG_PTLRPC_BODY_OFF 0
526 /* normal request/reply message record offset */
527 #define REQ_REC_OFF 1
528 #define REPLY_REC_OFF 1
530 /* ldlm request message body offset */
531 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
532 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
534 /* ldlm intent lock message body offset */
535 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
536 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
538 /* ldlm reply message body offset */
539 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
540 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
542 /* Flags that are operation-specific go in the top 16 bits. */
543 #define MSG_OP_FLAG_MASK 0xffff0000
544 #define MSG_OP_FLAG_SHIFT 16
546 /* Flags that apply to all requests are in the bottom 16 bits */
547 #define MSG_GEN_FLAG_MASK 0x0000ffff
548 #define MSG_LAST_REPLAY 0x0001
549 #define MSG_RESENT 0x0002
550 #define MSG_REPLAY 0x0004
551 /* #define MSG_AT_SUPPORT 0x0008
552 * This was used in early prototypes of adaptive timeouts, and while there
553 * shouldn't be any users of that code there also isn't a need for using this
554 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
555 #define MSG_REQ_REPLAY_DONE 0x0010
556 #define MSG_LOCK_REPLAY_DONE 0x0020
559 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
562 #define MSG_CONNECT_RECOVERING 0x1
563 #define MSG_CONNECT_RECONNECT 0x2
564 #define MSG_CONNECT_REPLAYABLE 0x4
565 //#define MSG_CONNECT_PEER 0x8
566 #define MSG_CONNECT_LIBCLIENT 0x10
567 #define MSG_CONNECT_INITIAL 0x20
568 #define MSG_CONNECT_ASYNC 0x40
569 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
570 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
573 #define OBD_CONNECT_RDONLY 0x00000001ULL /* client allowed read-only access */
574 #define OBD_CONNECT_INDEX 0x00000002ULL /* connect to specific LOV idx */
575 #define OBD_CONNECT_GRANT 0x00000008ULL /* OSC acquires grant at connect */
576 #define OBD_CONNECT_SRVLOCK 0x00000010ULL /* server takes locks for client */
577 #define OBD_CONNECT_VERSION 0x00000020ULL /* Server supports versions in ocd */
578 #define OBD_CONNECT_REQPORTAL 0x00000040ULL /* Separate portal for non-IO reqs */
579 #define OBD_CONNECT_ACL 0x00000080ULL /* client uses access control lists */
580 #define OBD_CONNECT_XATTR 0x00000100ULL /* client using extended attributes*/
581 #define OBD_CONNECT_TRUNCLOCK 0x00000400ULL /* locks on server for punch b=9528 */
582 #define OBD_CONNECT_IBITS 0x00001000ULL /* support for inodebits locks */
583 #define OBD_CONNECT_JOIN 0x00002000ULL /* files can be concatenated */
584 #define OBD_CONNECT_ATTRFID 0x00004000ULL /* Server supports GetAttr By Fid */
585 #define OBD_CONNECT_NODEVOH 0x00008000ULL /* No open handle for special nodes */
586 #define OBD_CONNECT_LCL_CLIENT 0x00010000ULL /* local 1.8 client */
587 #define OBD_CONNECT_RMT_CLIENT 0x00020000ULL /* Remote 1.8 client */
588 #define OBD_CONNECT_BRW_SIZE 0x00040000ULL /* Max bytes per rpc */
589 #define OBD_CONNECT_QUOTA64 0x00080000ULL /* 64bit qunit_data.qd_count b=10707*/
590 #define OBD_CONNECT_MDS_CAPA 0x00100000ULL /* MDS capability */
591 #define OBD_CONNECT_OSS_CAPA 0x00200000ULL /* OSS capability */
592 #define OBD_CONNECT_CANCELSET 0x00400000ULL /* Early batched cancels. */
593 #define OBD_CONNECT_SOM 0x00800000ULL /* SOM feature */
594 #define OBD_CONNECT_AT 0x01000000ULL /* client uses adaptive timeouts */
595 #define OBD_CONNECT_LRU_RESIZE 0x02000000ULL /* Lru resize feature. */
596 #define OBD_CONNECT_MDS_MDS 0x04000000ULL /* MDS-MDS connection*/
597 #define OBD_CONNECT_REAL 0x08000000ULL /* real connection */
598 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*shrink/enlarge qunit size
600 #define OBD_CONNECT_CKSUM 0x20000000ULL /* support several cksum algos */
601 #define OBD_CONNECT_FID 0x40000000ULL /* FID is supported by server */
603 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
604 * and lustre/utils/wirecheck.c */
606 #ifdef HAVE_LRU_RESIZE_SUPPORT
607 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
609 #define LRU_RESIZE_CONNECT_FLAG 0
612 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
613 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
614 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
615 OBD_CONNECT_NODEVOH |/* OBD_CONNECT_ATTRFID |*/\
616 OBD_CONNECT_LCL_CLIENT | \
617 OBD_CONNECT_RMT_CLIENT | \
618 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | \
619 OBD_CONNECT_MDS_MDS | OBD_CONNECT_CANCELSET | \
621 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_AT)
622 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
623 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
624 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
625 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
626 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET | \
627 OBD_CONNECT_CKSUM | LRU_RESIZE_CONNECT_FLAG | \
629 #define ECHO_CONNECT_SUPPORTED (0)
630 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT)
632 #define MAX_QUOTA_COUNT32 (0xffffffffULL)
634 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
635 ((patch)<<8) + (fix))
636 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
637 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
638 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
639 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
641 /* This structure is used for both request and reply.
643 * If we eventually have separate connect data for different types, which we
644 * almost certainly will, then perhaps we stick a union in here. */
645 struct obd_connect_data {
646 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
647 __u32 ocd_version; /* lustre release version number */
648 __u32 ocd_grant; /* initial cache grant amount (bytes) */
649 __u32 ocd_index; /* LOV index to connect to */
650 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
651 __u64 ocd_ibits_known; /* inode bits this client understands */
652 __u32 ocd_nllu; /* non-local-lustre-user */
653 __u32 ocd_nllg; /* non-local-lustre-group */
654 __u64 ocd_transno; /* first transno from client to be replayed */
655 __u32 ocd_group; /* MDS group on OST */
656 __u32 ocd_cksum_types; /* supported checksum algorithms */
657 __u64 padding1; /* also fix lustre_swab_connect */
658 __u64 padding2; /* also fix lustre_swab_connect */
661 extern void lustre_swab_connect(struct obd_connect_data *ocd);
664 * Supported checksum algorithms. Up to 32 checksum types are supported.
665 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
666 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
667 * algorithm and also the OBD_FL_CKSUM* flags.
670 OBD_CKSUM_CRC32 = 0x00000001,
671 OBD_CKSUM_ADLER = 0x00000002,
675 * OST requests: OBDO & OBD request records
680 OST_REPLY = 0, /* reply ? */
700 #define OST_FIRST_OPC OST_REPLY
702 typedef __u64 obd_id;
703 typedef __u64 obd_gr;
704 typedef __u64 obd_time;
705 typedef __u64 obd_size;
706 typedef __u64 obd_off;
707 typedef __u64 obd_blocks;
708 typedef __u32 obd_blksize;
709 typedef __u32 obd_mode;
710 typedef __u32 obd_uid;
711 typedef __u32 obd_gid;
712 typedef __u32 obd_flag;
713 typedef __u64 obd_valid;
714 typedef __u32 obd_count;
716 #define OBD_FL_INLINEDATA (0x00000001)
717 #define OBD_FL_OBDMDEXISTS (0x00000002)
718 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
719 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
720 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
721 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
722 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
723 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
724 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
727 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
728 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
731 #define OBD_FL_TRUNCLOCK (0x00000800)
736 #define OBD_FL_CKSUM_CRC32 (0x00001000)
737 #define OBD_FL_CKSUM_ADLER (0x00002000)
738 #define OBD_FL_CKSUM_ALL (OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER)
741 * This should not be smaller than sizeof(struct lustre_handle) + sizeof(struct
742 * llog_cookie) + sizeof(struct ll_fid). Nevertheless struct ll_fid is not
743 * longer stored in o_inline, we keep this just for case.
745 #define OBD_INLINESZ 80
747 /* Note: 64-bit types are 64-bit aligned in structure */
749 obd_valid o_valid; /* hot fields in this obdo */
753 obd_size o_size; /* o_size-o_blocks == ost_lvb */
757 obd_blocks o_blocks; /* brw: cli sent cached bytes */
760 /* 32-bit fields start here: keep an even number of them via padding */
761 obd_blksize o_blksize; /* optimal IO blocksize */
762 obd_mode o_mode; /* brw: cli sent cache remain */
766 obd_count o_nlink; /* brw: checksum */
767 obd_count o_generation;
768 obd_count o_misc; /* brw: o_dropped */
769 __u32 o_easize; /* epoch in ost writes */
771 __u32 o_stripe_idx; /* holds stripe idx */
773 char o_inline[OBD_INLINESZ];
774 /* lustre_handle + llog_cookie */
777 #define o_dirty o_blocks
778 #define o_undirty o_mode
779 #define o_dropped o_misc
780 #define o_cksum o_nlink
782 extern void lustre_swab_obdo (struct obdo *o);
785 #define LOV_MAGIC_V1 0x0BD10BD0
786 #define LOV_MAGIC LOV_MAGIC_V1
787 #define LOV_MAGIC_JOIN 0x0BD20BD0
789 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
790 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
791 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
792 #define LOV_PATTERN_CMOBD 0x200
794 #define LOV_OBJECT_GROUP_DEFAULT ~0ULL
795 #define LOV_OBJECT_GROUP_CLEAR 0ULL
797 #define lov_ost_data lov_ost_data_v1
798 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
799 __u64 l_object_id; /* OST object ID */
800 __u64 l_object_gr; /* OST object group (creating MDS number) */
801 __u32 l_ost_gen; /* generation of this l_ost_idx */
802 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
805 #define lov_mds_md lov_mds_md_v1
806 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
807 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
808 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
809 __u64 lmm_object_id; /* LOV object ID */
810 __u64 lmm_object_gr; /* LOV object group */
811 __u32 lmm_stripe_size; /* size of stripe in bytes */
812 __u32 lmm_stripe_count; /* num stripes in use for this object */
813 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
816 extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm);
818 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
819 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
821 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
822 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
823 #define XATTR_NAME_LOV "trusted.lov"
825 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
826 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
827 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
828 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
829 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
830 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
831 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
832 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
833 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
834 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
835 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
836 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
837 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
838 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
839 #define OBD_MD_FLINLINE (0x00008000ULL) /* inline data */
840 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
841 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
842 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
843 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file handle */
844 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
845 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
846 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
847 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
848 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
849 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
850 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
851 /* ->mds if epoch opens or closes */
852 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
853 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
854 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
855 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
856 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
858 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
859 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
860 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
862 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
863 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
864 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
865 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
866 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
867 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
868 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
869 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
870 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
872 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
873 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
874 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
875 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
877 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
878 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
879 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
880 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
881 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
883 static inline struct lustre_handle *obdo_handle(struct obdo *oa)
885 return (struct lustre_handle *)oa->o_inline;
888 static inline struct llog_cookie *obdo_logcookie(struct obdo *oa)
890 return (struct llog_cookie *)(oa->o_inline +
891 sizeof(struct lustre_handle));
893 /* don't forget obdo_fid which is way down at the bottom so it can
894 * come after the definition of llog_cookie */
907 __u32 os_state; /* positive error code on server */
919 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
920 #define OBD_STATFS_NODELAY 0x0001 /* requests should be send without delay
921 * and resends for avoid deadlocks */
923 /* ost_body.data values for OST_BRW */
925 #define OBD_BRW_READ 0x01
926 #define OBD_BRW_WRITE 0x02
927 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
928 #define OBD_BRW_SYNC 0x08
929 #define OBD_BRW_CHECK 0x10
930 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
931 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
932 #define OBD_BRW_DROP 0x80 /* drop the page after IO */
933 #define OBD_BRW_NOQUOTA 0x100
934 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
936 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
938 #define OST_MIN_PRECREATE 32
939 #define OST_MAX_PRECREATE 20000
948 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
950 /* multiple of 8 bytes => can array */
951 struct niobuf_remote {
957 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
959 /* request structure for OST's */
965 extern void lustre_swab_ost_body (struct ost_body *b);
966 extern void lustre_swab_ost_last_id(obd_id *id);
968 /* lock value block communicated between the filter and llite */
970 /* OST_LVB_ERR_INIT is needed because the return code in rc is
971 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
972 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
973 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
974 #define OST_LVB_IS_ERR(blocks) \
975 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
976 #define OST_LVB_SET_ERR(blocks, rc) \
977 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
978 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
988 extern void lustre_swab_ost_lvb(struct ost_lvb *);
997 MDS_GETATTR_NAME = 34,
1002 MDS_DISCONNECT = 39,
1008 MDS_DONE_WRITING = 45,
1010 MDS_QUOTACHECK = 47,
1013 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1019 #define MDS_FIRST_OPC MDS_GETATTR
1036 } mds_reint_t, mdt_reint_t;
1038 extern void lustre_swab_generic_32s (__u32 *val);
1040 /* the disposition of the intent outlines what was executed */
1041 #define DISP_IT_EXECD 0x00000001
1042 #define DISP_LOOKUP_EXECD 0x00000002
1043 #define DISP_LOOKUP_NEG 0x00000004
1044 #define DISP_LOOKUP_POS 0x00000008
1045 #define DISP_OPEN_CREATE 0x00000010
1046 #define DISP_OPEN_OPEN 0x00000020
1047 #define DISP_ENQ_COMPLETE 0x00400000
1048 #define DISP_ENQ_OPEN_REF 0x00800000
1049 #define DISP_ENQ_CREATE_REF 0x01000000
1050 #define DISP_OPEN_LOCK 0x02000000
1052 /* INODE LOCK PARTS */
1053 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1054 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1055 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1057 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
1058 #define MDS_INODELOCK_MAXSHIFT 2
1059 /* This FULL lock is useful to take on unlink sort of operations */
1060 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1063 __u64 id; /* holds object id */
1064 __u32 generation; /* holds object generation */
1066 __u32 f_type; /* holds object type or stripe idx when passing it to
1067 * OST for saving into EA. */
1070 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1072 #define MDS_STATUS_CONN 1
1073 #define MDS_STATUS_LOV 2
1075 struct mds_status_req {
1080 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
1082 /* mdt_thread_info.mti_flags. */
1084 /* The flag indicates Size-on-MDS attributes are changed. */
1085 MF_SOM_CHANGE = (1 << 0),
1086 /* Flags indicates an epoch opens or closes. */
1087 MF_EPOCH_OPEN = (1 << 1),
1088 MF_EPOCH_CLOSE = (1 << 2),
1089 MF_MDC_CANCEL_FID1 = (1 << 3),
1090 MF_MDC_CANCEL_FID2 = (1 << 4),
1091 MF_MDC_CANCEL_FID3 = (1 << 5),
1092 MF_MDC_CANCEL_FID4 = (1 << 6),
1095 #define MF_SOM_LOCAL_FLAGS (MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID2 | \
1096 MF_MDC_CANCEL_FID3 | MF_MDC_CANCEL_FID4)
1098 #define MDS_BFLAG_UNCOMMITTED_WRITES 0x1
1099 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
1101 /* these should be identical to their EXT3_*_FL counterparts, and are
1102 * redefined here only to avoid dragging in ext3_fs.h */
1103 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
1104 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
1105 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
1106 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
1107 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1110 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
1111 * and we need to decode these into local S_* flags in the inode. Otherwise
1112 * we pass flags straight through (see bug 9486). */
1113 static inline int ll_ext_to_inode_flags(int flags)
1115 return (flags & MDS_BFLAG_EXT_FLAGS) ?
1116 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
1117 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
1118 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
1119 #if defined(S_DIRSYNC)
1120 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1122 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1123 (flags & ~MDS_BFLAG_EXT_FLAGS);
1126 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1127 * and we pass these straight through. Otherwise we need to convert from
1128 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1129 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1131 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1132 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1133 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1134 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1135 #if defined(S_DIRSYNC)
1136 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1138 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1145 struct lustre_handle handle;
1147 __u64 size; /* Offset, in the case of MDS_READPAGE */
1151 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1153 __u64 ino; /* for 1.6 compatibility */
1160 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1162 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1163 __u32 generation; /* for 1.6 compatibility */
1168 __u32 max_cookiesize;
1169 __u32 padding_4; /* also fix lustre_swab_mdt_body */
1175 struct lustre_handle handle;
1177 __u64 size; /* Offset, in the case of MDS_READPAGE */
1181 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1190 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1192 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1198 __u32 max_cookiesize;
1199 __u32 padding_4; /* also fix lustre_swab_mds_body */
1202 extern void lustre_swab_mds_body (struct mds_body *b);
1203 extern void lustre_swab_mdt_body (struct mdt_body *b);
1206 struct lustre_handle handle;
1212 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1214 #define Q_QUOTACHECK 0x800100
1215 #define Q_INITQUOTA 0x800101 /* init slave limits */
1216 #define Q_GETOINFO 0x800102 /* get obd quota info */
1217 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1219 #define Q_TYPESET(oqc, type) \
1220 ((oqc)->qc_type == type || (oqc)->qc_type == UGQUOTA)
1222 #define Q_GETOCMD(oqc) \
1223 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1225 struct obd_quotactl {
1230 struct obd_dqinfo qc_dqinfo;
1231 struct obd_dqblk qc_dqblk;
1234 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1236 /* inode access permission for remote user, the inode info are omitted,
1237 * for client knows them. */
1238 struct mds_remote_perm {
1243 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1246 /* permissions for md_perm.mp_perm */
1248 CFS_SETUID_PERM = 0x01,
1249 CFS_SETGID_PERM = 0x02,
1250 CFS_SETGRP_PERM = 0x04,
1251 CFS_RMTACL_PERM = 0x08
1254 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1256 struct mdt_remote_perm {
1261 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1264 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1266 struct mds_rec_setattr {
1273 struct ll_fid sa_fid;
1274 __u64 sa_valid; /* MDS_ATTR_* attributes */
1281 __u32 sa_attr_flags;
1282 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1285 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1287 struct mdt_rec_setattr {
1294 struct lu_fid sa_fid;
1303 __u32 sa_attr_flags;
1311 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1314 * Attribute flags used in mds_rec_setattr::sa_valid.
1315 * The kernel's #defines for ATTR_* should not be used over the network
1316 * since the client and MDS may run different kernels (see bug 13828)
1317 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1319 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1320 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1321 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1322 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1323 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1324 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1325 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1326 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1327 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1328 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1329 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1330 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1331 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1332 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1333 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1334 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1337 #define FMODE_READ 00000001
1338 #define FMODE_WRITE 00000002
1341 #define FMODE_EPOCH 01000000
1342 #define FMODE_EPOCHLCK 02000000
1343 #define FMODE_SOM 04000000
1344 #define FMODE_CLOSED 0
1346 #define MDS_OPEN_CREATED 00000010
1347 #define MDS_OPEN_CROSS 00000020
1349 #define MDS_FMODE_EXEC 00000004
1350 #define MDS_OPEN_CREAT 00000100
1351 #define MDS_OPEN_EXCL 00000200
1352 #define MDS_OPEN_TRUNC 00001000
1353 #define MDS_OPEN_APPEND 00002000
1354 #define MDS_OPEN_SYNC 00010000
1355 #define MDS_OPEN_DIRECTORY 00200000
1357 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1358 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1359 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1360 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1361 * with default ACL */
1362 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1363 * actually, this is for create, not
1364 * conflict with other open flags */
1365 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1366 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1367 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1369 /* permission for create non-directory file */
1370 #define MAY_CREATE (1 << 7)
1371 /* permission for create directory file */
1372 #define MAY_LINK (1 << 8)
1373 /* permission for delete from the directory */
1374 #define MAY_UNLINK (1 << 9)
1375 /* source's permission for rename */
1376 #define MAY_RENAME_SRC (1 << 10)
1377 /* target's permission for rename */
1378 #define MAY_RENAME_TAR (1 << 11)
1379 /* part (parent's) VTX permission check */
1380 #define MAY_VTX_PART (1 << 12)
1381 /* full VTX permission check */
1382 #define MAY_VTX_FULL (1 << 13)
1383 /* lfs rgetfacl permission check */
1384 #define MAY_RGETFACL (1 << 14)
1387 MDS_CHECK_SPLIT = 1 << 0,
1388 MDS_CROSS_REF = 1 << 1,
1389 MDS_VTX_BYPASS = 1 << 2,
1390 MDS_PERM_BYPASS = 1 << 3,
1394 struct mds_rec_join {
1395 struct ll_fid jr_fid;
1399 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1401 struct mdt_rec_join {
1402 struct lu_fid jr_fid;
1406 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1408 struct mds_rec_create {
1413 __u32 cr_flags; /* for use with open */
1415 struct ll_fid cr_fid;
1416 struct ll_fid cr_replayfid;
1420 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1421 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1422 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1423 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1424 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1427 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1429 struct mdt_rec_create {
1436 struct lu_fid cr_fid1;
1437 struct lu_fid cr_fid2;
1438 struct lustre_handle cr_old_handle; /* u64 handle in case of open replay */
1442 __u64 cr_padding_1; /* pad for 64 bits*/
1445 __u32 cr_flags; /* for use with open */
1451 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1453 struct mds_rec_link {
1460 struct ll_fid lk_fid1;
1461 struct ll_fid lk_fid2;
1463 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1464 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1465 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1466 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1469 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1471 struct mdt_rec_link {
1478 struct lu_fid lk_fid1;
1479 struct lu_fid lk_fid2;
1493 struct mds_rec_unlink {
1500 struct ll_fid ul_fid1;
1501 struct ll_fid ul_fid2;
1503 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1504 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1505 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1506 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1509 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1511 struct mdt_rec_unlink {
1518 struct lu_fid ul_fid1;
1519 struct lu_fid ul_fid2;
1533 struct mds_rec_rename {
1540 struct ll_fid rn_fid1;
1541 struct ll_fid rn_fid2;
1543 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1544 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1545 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1546 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1549 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1551 struct mdt_rec_rename {
1558 struct lu_fid rn_fid1;
1559 struct lu_fid rn_fid2;
1565 __u32 rn_bias; /* some operation flags */
1566 __u32 rn_mode; /* cross-ref rename has mode */
1573 struct mdt_rec_setxattr {
1580 struct lu_fid sx_fid;
1581 __u64 sx_padding_1; /* These three members are lu_fid size */
1593 __u32 sx_padding_10;
1594 __u32 sx_padding_11;
1597 struct mdt_rec_reint {
1604 struct lu_fid rr_fid1;
1605 struct lu_fid rr_fid2;
1613 __u32 rr_padding_1; /* also fix lustre_swab_mdt_rec_reint */
1614 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
1615 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
1616 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1619 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
1622 __u32 ld_tgt_count; /* how many MDS's */
1623 __u32 ld_active_tgt_count; /* how many active */
1624 struct obd_uuid ld_uuid;
1627 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1634 extern void lustre_swab_md_fld (struct md_fld *mf);
1639 FLD_FIRST_OPC = FLD_QUERY
1645 SEQ_FIRST_OPC = SEQ_QUERY
1649 SEQ_ALLOC_SUPER = 0,
1654 * LOV data structures
1657 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1658 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1659 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
1661 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1662 /* The size of the buffer the lov/mdc reserves for the
1663 * array of UUIDs returned by the MDS. With the current
1664 * protocol, this will limit the max number of OSTs per LOV */
1666 #define LOV_DESC_MAGIC 0xB0CCDE5C
1668 /* LOV settings descriptor (should only contain static info) */
1670 __u32 ld_tgt_count; /* how many OBD's */
1671 __u32 ld_active_tgt_count; /* how many active */
1672 __u32 ld_default_stripe_count; /* how many objects are used */
1673 __u32 ld_pattern; /* default PATTERN_RAID0 */
1674 __u64 ld_default_stripe_size; /* in bytes */
1675 __u64 ld_default_stripe_offset; /* in bytes */
1676 __u32 ld_padding_0; /* unused */
1677 __u32 ld_qos_maxage; /* in second */
1678 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1679 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1680 struct obd_uuid ld_uuid;
1683 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1685 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1690 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1695 LDLM_BL_CALLBACK = 104,
1696 LDLM_CP_CALLBACK = 105,
1697 LDLM_GL_CALLBACK = 106,
1700 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1702 #define RES_NAME_SIZE 4
1703 struct ldlm_res_id {
1704 __u64 name[RES_NAME_SIZE];
1707 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1722 #define LCK_MODE_NUM 7
1732 #define LDLM_MIN_TYPE LDLM_PLAIN
1734 struct ldlm_extent {
1740 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
1741 struct ldlm_extent *ex2)
1743 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
1746 struct ldlm_inodebits {
1753 __u64 blocking_export; /* not actually used over the wire */
1754 __u32 blocking_pid; /* not actually used over the wire */
1758 /* it's important that the fields of the ldlm_extent structure match
1759 * the first fields of the ldlm_flock structure because there is only
1760 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1761 * this ever changes we will need to swab the union differently based
1762 * on the resource type. */
1765 struct ldlm_extent l_extent;
1766 struct ldlm_flock l_flock;
1767 struct ldlm_inodebits l_inodebits;
1768 } ldlm_policy_data_t;
1770 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1772 struct ldlm_intent {
1776 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1778 struct ldlm_resource_desc {
1779 ldlm_type_t lr_type;
1780 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1781 struct ldlm_res_id lr_name;
1784 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1786 struct ldlm_lock_desc {
1787 struct ldlm_resource_desc l_resource;
1788 ldlm_mode_t l_req_mode;
1789 ldlm_mode_t l_granted_mode;
1790 ldlm_policy_data_t l_policy_data;
1793 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1795 #define LDLM_LOCKREQ_HANDLES 2
1796 #define LDLM_ENQUEUE_CANCEL_OFF 1
1798 struct ldlm_request {
1801 struct ldlm_lock_desc lock_desc;
1802 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
1805 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1807 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
1808 * Otherwise, 2 are available. */
1809 #define ldlm_request_bufsize(count,type) \
1811 int _avail = LDLM_LOCKREQ_HANDLES; \
1812 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
1813 sizeof(struct ldlm_request) + \
1814 (count > _avail ? count - _avail : 0) * \
1815 sizeof(struct lustre_handle); \
1820 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1821 struct ldlm_lock_desc lock_desc;
1822 struct lustre_handle lock_handle;
1823 __u64 lock_policy_res1;
1824 __u64 lock_policy_res2;
1827 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1830 * Opcodes for mountconf (mgs and mgc)
1835 MGS_EXCEPTION, /* node died, etc. */
1836 MGS_TARGET_REG, /* whenever target starts up */
1841 #define MGS_FIRST_OPC MGS_CONNECT
1843 #define MGS_PARAM_MAXLEN 1024
1844 #define KEY_SET_INFO "set_info"
1846 struct mgs_send_param {
1847 char mgs_param[MGS_PARAM_MAXLEN];
1850 /* We pass this info to the MGS so it can write config logs */
1851 #define MTI_NAME_MAXLEN 64
1852 #define MTI_PARAM_MAXLEN 4096
1853 #define MTI_NIDS_MAX 32
1854 struct mgs_target_info {
1855 __u32 mti_lustre_ver;
1856 __u32 mti_stripe_index;
1857 __u32 mti_config_ver;
1859 __u32 mti_nid_count;
1860 __u32 padding; /* 64 bit align */
1861 char mti_fsname[MTI_NAME_MAXLEN];
1862 char mti_svname[MTI_NAME_MAXLEN];
1863 char mti_uuid[sizeof(struct obd_uuid)];
1864 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
1865 char mti_params[MTI_PARAM_MAXLEN];
1868 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
1870 /* Config marker flags (in config log) */
1871 #define CM_START 0x01
1873 #define CM_SKIP 0x04
1874 #define CM_UPGRADE146 0x08
1875 #define CM_EXCLUDE 0x10
1876 #define CM_START_SKIP (CM_START | CM_SKIP)
1879 __u32 cm_step; /* aka config version */
1881 __u32 cm_vers; /* lustre release version number */
1882 __u32 padding; /* 64 bit align */
1883 time_t cm_createtime; /*when this record was first created */
1884 time_t cm_canceltime; /*when this record is no longer valid*/
1885 char cm_tgtname[MTI_NAME_MAXLEN];
1886 char cm_comment[MTI_NAME_MAXLEN];
1890 * Opcodes for multiple servers.
1899 #define OBD_FIRST_OPC OBD_PING
1901 /* catalog of log objects */
1903 /* Identifier for a single log object */
1908 } __attribute__((packed));
1910 /* Records written to the CATALOGS list */
1911 #define CATLIST "CATALOGS"
1913 struct llog_logid lci_logid;
1917 } __attribute__((packed));
1919 /*join file lov mds md*/
1920 struct lov_mds_md_join {
1921 struct lov_mds_md lmmj_md;
1922 /*join private info*/
1923 struct llog_logid lmmj_array_id; /*array object id*/
1924 __u32 lmmj_extent_count; /*array extent count*/
1927 /* Log data record types - there is no specific reason that these need to
1928 * be related to the RPC opcodes, but no reason not to (may be handy later?)
1930 #define LLOG_OP_MAGIC 0x10600000
1931 #define LLOG_OP_MASK 0xfff00000
1934 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
1935 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
1936 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
1937 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
1938 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
1939 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
1940 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
1941 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
1942 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
1943 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
1944 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
1948 * for now, continue to support old pad records which have 0 for their
1949 * type but still need to be swabbed for their length
1951 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
1952 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
1953 __swab32(LLOG_OP_MAGIC) || \
1954 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
1956 /* Log record header - stored in little endian order.
1957 * Each record must start with this struct, end with a llog_rec_tail,
1958 * and be a multiple of 256 bits in size.
1960 struct llog_rec_hdr {
1967 struct llog_rec_tail {
1972 struct llog_logid_rec {
1973 struct llog_rec_hdr lid_hdr;
1974 struct llog_logid lid_id;
1980 struct llog_rec_tail lid_tail;
1981 } __attribute__((packed));
1983 /* MDS extent description
1984 * It is for joined file extent info, each extent info for joined file
1985 * just like (start, end, lmm).
1987 struct mds_extent_desc {
1988 __u64 med_start; /* extent start */
1989 __u64 med_len; /* extent length */
1990 struct lov_mds_md med_lmm; /* extent's lmm */
1992 /*Joined file array extent log record*/
1993 struct llog_array_rec {
1994 struct llog_rec_hdr lmr_hdr;
1995 struct mds_extent_desc lmr_med;
1996 struct llog_rec_tail lmr_tail;
1999 struct llog_create_rec {
2000 struct llog_rec_hdr lcr_hdr;
2001 struct ll_fid lcr_fid;
2005 struct llog_rec_tail lcr_tail;
2006 } __attribute__((packed));
2008 struct llog_orphan_rec {
2009 struct llog_rec_hdr lor_hdr;
2013 struct llog_rec_tail lor_tail;
2014 } __attribute__((packed));
2016 struct llog_unlink_rec {
2017 struct llog_rec_hdr lur_hdr;
2021 struct llog_rec_tail lur_tail;
2022 } __attribute__((packed));
2024 struct llog_setattr_rec {
2025 struct llog_rec_hdr lsr_hdr;
2031 struct llog_rec_tail lsr_tail;
2032 } __attribute__((packed));
2034 struct llog_size_change_rec {
2035 struct llog_rec_hdr lsc_hdr;
2036 struct ll_fid lsc_fid;
2039 struct llog_rec_tail lsc_tail;
2040 } __attribute__((packed));
2045 } __attribute__((packed));
2047 struct llog_gen_rec {
2048 struct llog_rec_hdr lgr_hdr;
2049 struct llog_gen lgr_gen;
2050 struct llog_rec_tail lgr_tail;
2052 /* On-disk header structure of each log object, stored in little endian order */
2053 #define LLOG_CHUNK_SIZE 8192
2054 #define LLOG_HEADER_SIZE (96)
2055 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2057 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2059 /* flags for the logs */
2060 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
2061 #define LLOG_F_IS_CAT 0x2
2062 #define LLOG_F_IS_PLAIN 0x4
2064 struct llog_log_hdr {
2065 struct llog_rec_hdr llh_hdr;
2066 __u64 llh_timestamp;
2068 __u32 llh_bitmap_offset;
2072 /* for a catalog the first plain slot is next to it */
2073 struct obd_uuid llh_tgtuuid;
2074 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2075 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2076 struct llog_rec_tail llh_tail;
2077 } __attribute__((packed));
2079 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
2080 llh->llh_bitmap_offset - \
2081 sizeof(llh->llh_tail)) * 8)
2083 /* log cookies are used to reference a specific log file and a record therein */
2084 struct llog_cookie {
2085 struct llog_logid lgc_lgl;
2089 } __attribute__((packed));
2092 enum llogd_rpc_ops {
2093 LLOG_ORIGIN_HANDLE_CREATE = 501,
2094 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2095 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2096 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2097 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2098 LLOG_ORIGIN_CONNECT = 506,
2099 LLOG_CATINFO = 507, /* for lfs catinfo */
2100 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2101 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2103 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2107 struct llog_logid lgd_logid;
2109 __u32 lgd_llh_flags;
2111 __u32 lgd_saved_index;
2113 __u64 lgd_cur_offset;
2114 } __attribute__((packed));
2116 struct llogd_conn_body {
2117 struct llog_gen lgdc_gen;
2118 struct llog_logid lgdc_logid;
2119 __u32 lgdc_ctxt_idx;
2120 } __attribute__((packed));
2122 struct lov_user_ost_data_join { /* per-stripe data structure */
2123 __u64 l_extent_start; /* extent start*/
2124 __u64 l_extent_end; /* extent end*/
2125 __u64 l_object_id; /* OST object ID */
2126 __u64 l_object_gr; /* OST object group (creating MDS number) */
2127 __u32 l_ost_gen; /* generation of this OST index */
2128 __u32 l_ost_idx; /* OST index in LOV */
2129 } __attribute__((packed));
2131 struct lov_user_md_join { /* LOV EA user data (host-endian) */
2132 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
2133 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
2134 __u64 lmm_object_id; /* LOV object ID */
2135 __u64 lmm_object_gr; /* LOV object group */
2136 __u32 lmm_stripe_size; /* size of stripe in bytes */
2137 __u32 lmm_stripe_count; /* num stripes in use for this object */
2138 __u32 lmm_extent_count; /* extent count of lmm*/
2139 __u64 lmm_tree_id; /* mds tree object id */
2140 __u64 lmm_tree_gen; /* mds tree object gen */
2141 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
2142 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
2143 } __attribute__((packed));
2145 extern void lustre_swab_lov_user_md(struct lov_user_md *lum);
2146 extern void lustre_swab_lov_user_md_objects(struct lov_user_md *lum);
2147 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
2150 extern void lustre_swab_llogd_body (struct llogd_body *d);
2151 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
2152 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
2153 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
2154 struct llog_rec_tail *tail);
2157 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
2159 /* quota. fixed by tianzy for bug10707 */
2160 #define QUOTA_IS_GRP 0X1UL /* 0 is user, 1 is group. Used by qd_flags*/
2161 #define QUOTA_IS_BLOCK 0x2UL /* 0 is inode, 1 is block. Used by qd_flags*/
2164 __u32 qd_id; /* ID appiles to (uid, gid) */
2165 __u32 qd_flags; /* Quota type (USRQUOTA, GRPQUOTA) occupy one bit;
2166 * Block quota or file quota occupy one bit */
2167 __u64 qd_count; /* acquire/release count (bytes for block quota) */
2170 struct qunit_data_old {
2171 __u32 qd_id; /* ID appiles to (uid, gid) */
2172 __u32 qd_type; /* Quota type (USRQUOTA, GRPQUOTA) */
2173 __u32 qd_count; /* acquire/release count (bytes for block quota) */
2174 __u32 qd_isblk; /* Block quota or file quota */
2177 extern void lustre_swab_qdata(struct qunit_data *d);
2178 extern void lustre_swab_qdata_old(struct qunit_data_old *d);
2179 extern struct qunit_data *lustre_quota_old_to_new(struct qunit_data_old *d);
2180 extern struct qunit_data_old *lustre_quota_new_to_old(struct qunit_data *d);
2187 #define JOIN_FILE_ALIGN 4096
2189 /* security opcodes */
2192 SEC_CTX_INIT_CONT = 802,
2195 SEC_FIRST_OPC = SEC_CTX_INIT
2199 * capa related definitions
2201 #define CAPA_HMAC_MAX_LEN 64
2202 #define CAPA_HMAC_KEY_MAX_LEN 56
2204 /* NB take care when changing the sequence of elements this struct,
2205 * because the offset info is used in find_capa() */
2206 struct lustre_capa {
2207 struct lu_fid lc_fid; /* fid */
2208 __u64 lc_opc; /* operations allowed */
2209 __u32 lc_uid; /* uid, it is obsolete, but maybe used in
2210 * future, reserve it for 64-bits aligned.*/
2211 __u32 lc_flags; /* HMAC algorithm & flags */
2212 __u32 lc_keyid; /* key used for the capability */
2213 __u32 lc_timeout; /* capa timeout value (sec) */
2214 __u64 lc_expiry; /* expiry time (sec) */
2215 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /* HMAC */
2216 } __attribute__((packed));
2218 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
2220 /* lustre_capa.lc_opc */
2222 CAPA_OPC_BODY_WRITE = 1<<0, /* write object data */
2223 CAPA_OPC_BODY_READ = 1<<1, /* read object data */
2224 CAPA_OPC_INDEX_LOOKUP = 1<<2, /* lookup object fid */
2225 CAPA_OPC_INDEX_INSERT = 1<<3, /* insert object fid */
2226 CAPA_OPC_INDEX_DELETE = 1<<4, /* delete object fid */
2227 CAPA_OPC_OSS_WRITE = 1<<5, /* write oss object data */
2228 CAPA_OPC_OSS_READ = 1<<6, /* read oss object data */
2229 CAPA_OPC_OSS_TRUNC = 1<<7, /* truncate oss object */
2230 CAPA_OPC_META_WRITE = 1<<8, /* write object meta data */
2231 CAPA_OPC_META_READ = 1<<9, /* read object meta data */
2235 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2236 #define CAPA_OPC_MDS_ONLY \
2237 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2238 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2239 #define CAPA_OPC_OSS_ONLY \
2240 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC)
2241 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2242 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2244 /* MDS capability covers object capability for operations of body r/w
2245 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
2246 * while OSS capability only covers object capability for operations of
2247 * oss data(file content) r/w/truncate.
2249 static inline int capa_for_mds(struct lustre_capa *c)
2251 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2254 static inline int capa_for_oss(struct lustre_capa *c)
2256 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2259 /* lustre_capa.lc_hmac_alg */
2261 CAPA_HMAC_ALG_SHA1 = 1, /* sha1 algorithm */
2265 #define CAPA_FL_MASK 0x00ffffff
2266 #define CAPA_HMAC_ALG_MASK 0xff000000
2268 struct lustre_capa_key {
2269 __u64 lk_mdsid; /* mds# */
2270 __u32 lk_keyid; /* key# */
2272 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /* key */
2273 } __attribute__((packed));
2275 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2277 /* quota check function */
2278 #define QUOTA_RET_OK 0 /* return successfully */
2279 #define QUOTA_RET_NOQUOTA 1 /* not support quota */
2280 #define QUOTA_RET_NOLIMIT 2 /* quota limit isn't set */
2281 #define QUOTA_RET_ACQUOTA 3 /* need to acquire extra quota */