1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/include/lustre/lustre_idl.h
38 * Lustre wire protocol definitions.
41 /** \defgroup lustreidl lustreidl
43 * Lustre wire protocol definitions.
45 * We assume all nodes are either little-endian or big-endian, and we
46 * always send messages in the sender's native format. The receiver
47 * detects the message format by checking the 'magic' field of the message
48 * (see lustre_msg_swabbed() below).
49 * ALL structs passing over the wire should be declared here. Structs
50 * that are used in interfaces with userspace should go in lustre_user.h.
52 * All structs being declared here should be built from simple fixed-size
53 * types (__u8, __u16, __u32, __u64) or be built from other types or
54 * structs also declared in this file. Similarly, all flags and magic
55 * values in those structs should also be declared here. This ensures
56 * that the Lustre wire protocol is not influenced by external dependencies.
58 * The only other acceptable items in this file are VERY SIMPLE accessor
59 * functions to avoid callers grubbing inside the structures, and the
60 * prototypes of the swabber functions for each struct. Nothing that
61 * depends on external functions or definitions should be in here.
63 * Structs must be properly aligned to put 64-bit values on an 8-byte
64 * boundary. Any structs being added here must also be added to
65 * utils/wirecheck.c and "make newwiretest" run to regenerate the
66 * utils/wiretest.c sources. This allows us to verify that wire structs
67 * have the proper alignment/size on all architectures.
69 * DO NOT CHANGE any of the structs, flags, values declared here and used
70 * in released Lustre versions. Some structs may have padding fields that
71 * can be used. Some structs might allow addition at the end (verify this
72 * in the code to ensure that new/old clients that see this larger struct
73 * do not fail, otherwise you need to implement protocol compatibility).
75 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
76 * implemented either here, inline (trivial implementations) or in
77 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
78 * endian, in-place in the message buffer.
80 * A swabber takes a single pointer argument. The caller must already have
81 * verified that the length of the message buffer >= sizeof (type).
83 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
84 * may be defined that swabs just the variable part, after the caller has
85 * verified that the message buffer is large enough.
90 #ifndef _LUSTRE_IDL_H_
91 #define _LUSTRE_IDL_H_
93 #include <libcfs/libcfs.h>
95 /* Defn's shared with user-space. */
96 #include <lustre/lustre_user.h>
97 #include <lustre/ll_fiemap.h>
102 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
103 * FOO_REPLY_PORTAL is for incoming replies on the FOO
104 * FOO_BULK_PORTAL is for incoming bulk on the FOO
107 #define CONNMGR_REQUEST_PORTAL 1
108 #define CONNMGR_REPLY_PORTAL 2
109 //#define OSC_REQUEST_PORTAL 3
110 #define OSC_REPLY_PORTAL 4
111 //#define OSC_BULK_PORTAL 5
112 #define OST_IO_PORTAL 6
113 #define OST_CREATE_PORTAL 7
114 #define OST_BULK_PORTAL 8
115 //#define MDC_REQUEST_PORTAL 9
116 #define MDC_REPLY_PORTAL 10
117 //#define MDC_BULK_PORTAL 11
118 #define MDS_REQUEST_PORTAL 12
119 //#define MDS_REPLY_PORTAL 13
120 #define MDS_BULK_PORTAL 14
121 #define LDLM_CB_REQUEST_PORTAL 15
122 #define LDLM_CB_REPLY_PORTAL 16
123 #define LDLM_CANCEL_REQUEST_PORTAL 17
124 #define LDLM_CANCEL_REPLY_PORTAL 18
125 //#define PTLBD_REQUEST_PORTAL 19
126 //#define PTLBD_REPLY_PORTAL 20
127 //#define PTLBD_BULK_PORTAL 21
128 #define MDS_SETATTR_PORTAL 22
129 #define MDS_READPAGE_PORTAL 23
130 #define MDS_MDS_PORTAL 24
132 #define MGC_REPLY_PORTAL 25
133 #define MGS_REQUEST_PORTAL 26
134 #define MGS_REPLY_PORTAL 27
135 #define OST_REQUEST_PORTAL 28
136 #define FLD_REQUEST_PORTAL 29
137 #define SEQ_METADATA_PORTAL 30
138 #define SEQ_DATA_PORTAL 31
139 #define SEQ_CONTROLLER_PORTAL 32
144 #define SVC_RUNNING 8
145 #define SVC_STOPPING 16
146 #define SVC_STOPPED 32
149 #define PTL_RPC_MSG_REQUEST 4711
150 #define PTL_RPC_MSG_ERR 4712
151 #define PTL_RPC_MSG_REPLY 4713
153 /* DON'T use swabbed values of MAGIC as magic! */
154 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
155 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
157 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
158 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
160 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
162 #define PTLRPC_MSG_VERSION 0x00000003
163 #define LUSTRE_VERSION_MASK 0xffff0000
164 #define LUSTRE_OBD_VERSION 0x00010000
165 #define LUSTRE_MDS_VERSION 0x00020000
166 #define LUSTRE_OST_VERSION 0x00030000
167 #define LUSTRE_DLM_VERSION 0x00040000
168 #define LUSTRE_LOG_VERSION 0x00050000
169 #define LUSTRE_MGS_VERSION 0x00060000
171 typedef __u64 mdsno_t;
172 typedef __u64 seqno_t;
179 static inline __u64 range_space(struct lu_range *r)
181 return r->lr_end - r->lr_start;
184 static inline void range_zero(struct lu_range *r)
186 r->lr_start = r->lr_end = 0;
189 static inline int range_within(struct lu_range *r,
192 return s >= r->lr_start && s < r->lr_end;
195 static inline void range_alloc(struct lu_range *r,
199 r->lr_start = s->lr_start;
200 r->lr_end = s->lr_start + w;
204 static inline int range_is_sane(struct lu_range *r)
206 return (r->lr_end >= r->lr_start);
209 static inline int range_is_zero(struct lu_range *r)
211 return (r->lr_start == 0 && r->lr_end == 0);
214 static inline int range_is_exhausted(struct lu_range *r)
216 return range_space(r) == 0;
219 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x]"
221 #define PRANGE(range) \
225 /** \defgroup lu_fid lu_fid
231 * Fid is a cluster-wide unique identifier of a file or an object
232 * (stripe). Fids are never reused. Fids are transmitted across network (in
233 * the sender byte-ordering), and stored on disk in a packed form (struct
234 * lu_fid_pack) in a big-endian order.
238 * fid sequence. Sequence is a unit of migration: all files (objects)
239 * with fids from a given sequence are stored on the same
242 * Lustre should support 2 ^ 64 objects, thus even if one
243 * sequence has one object we will never reach this value.
246 /** fid number within sequence. */
249 * fid version, used to distinguish different versions (in the sense
250 * of snapshots, etc.) of the same file system object. Not currently
260 /* initial fid id value */
261 LUSTRE_FID_INIT_OID = 1UL
264 /** returns fid object sequence */
265 static inline __u64 fid_seq(const struct lu_fid *fid)
270 /** returns fid object id */
271 static inline __u32 fid_oid(const struct lu_fid *fid)
276 /** returns fid object version */
277 static inline __u32 fid_ver(const struct lu_fid *fid)
282 static inline void fid_zero(struct lu_fid *fid)
284 memset(fid, 0, sizeof(*fid));
287 /* Normal FID sequence starts from this value, i.e. 1<<33 */
288 #define FID_SEQ_START 0x200000000ULL
290 /* IDIF sequence starts from this value, i.e. 1<<32 */
291 #define IDIF_SEQ_START 0x100000000ULL
294 * Check if a fid is igif or not.
295 * \param fid the fid to be tested.
296 * \return true if the fid is a igif; otherwise false.
298 static inline int fid_is_igif(const struct lu_fid *fid)
300 return fid_seq(fid) > 0 && fid_seq(fid) < IDIF_SEQ_START;
304 * Check if a fid is idif or not.
305 * \param fid the fid to be tested.
306 * \return true if the fid is a idif; otherwise false.
308 static inline int fid_is_idif(const struct lu_fid *fid)
310 return fid_seq(fid) >= IDIF_SEQ_START && fid_seq(fid) < FID_SEQ_START;
314 * Get inode number from a igif.
315 * \param fid a igif to get inode number from.
316 * \return inode number for the igif.
318 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
324 * Get inode generation from a igif.
325 * \param fid a igif to get inode generation from.
326 * \return inode generation for the igif.
328 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
333 #define DFID "[0x%16.16"LPF64"x/0x%8.8x:0x%8.8x]"
340 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
342 /* check that all fields are converted */
343 CLASSERT(sizeof *src ==
344 sizeof fid_seq(src) +
345 sizeof fid_oid(src) + sizeof fid_ver(src));
346 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
347 dst->f_seq = cpu_to_le64(fid_seq(src));
348 dst->f_oid = cpu_to_le32(fid_oid(src));
349 dst->f_ver = cpu_to_le32(fid_ver(src));
352 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
354 /* check that all fields are converted */
355 CLASSERT(sizeof *src ==
356 sizeof fid_seq(src) +
357 sizeof fid_oid(src) + sizeof fid_ver(src));
358 dst->f_seq = le64_to_cpu(fid_seq(src));
359 dst->f_oid = le32_to_cpu(fid_oid(src));
360 dst->f_ver = le32_to_cpu(fid_ver(src));
361 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
364 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
366 /* check that all fields are converted */
367 CLASSERT(sizeof *src ==
368 sizeof fid_seq(src) +
369 sizeof fid_oid(src) + sizeof fid_ver(src));
370 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
371 dst->f_seq = cpu_to_be64(fid_seq(src));
372 dst->f_oid = cpu_to_be32(fid_oid(src));
373 dst->f_ver = cpu_to_be32(fid_ver(src));
376 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
378 /* check that all fields are converted */
379 CLASSERT(sizeof *src ==
380 sizeof fid_seq(src) +
381 sizeof fid_oid(src) + sizeof fid_ver(src));
382 dst->f_seq = be64_to_cpu(fid_seq(src));
383 dst->f_oid = be32_to_cpu(fid_oid(src));
384 dst->f_ver = be32_to_cpu(fid_ver(src));
385 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
389 * Storage representation for fids.
391 * Variable size, first byte contains the length of the whole record.
395 char fp_area[sizeof(struct lu_fid)];
398 void fid_pack(struct lu_fid_pack *pack, const struct lu_fid *fid,
399 struct lu_fid *befider);
400 int fid_unpack(const struct lu_fid_pack *pack, struct lu_fid *fid);
402 static inline int fid_is_sane(const struct lu_fid *fid)
406 ((fid_seq(fid) >= FID_SEQ_START && fid_oid(fid) != 0
407 && fid_ver(fid) == 0) ||
411 static inline int fid_is_zero(const struct lu_fid *fid)
413 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
416 extern void lustre_swab_lu_fid(struct lu_fid *fid);
417 extern void lustre_swab_lu_range(struct lu_range *range);
419 static inline int lu_fid_eq(const struct lu_fid *f0,
420 const struct lu_fid *f1)
422 /* Check that there is no alignment padding. */
423 CLASSERT(sizeof *f0 ==
424 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
425 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
426 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
427 return memcmp(f0, f1, sizeof *f0) == 0;
430 #define __diff_normalize(val0, val1) \
432 typeof(val0) __val0 = (val0); \
433 typeof(val1) __val1 = (val1); \
435 (__val0 == __val1 ? 0 : __val0 > __val1 ? +1 : -1); \
438 static inline int lu_fid_cmp(const struct lu_fid *f0,
439 const struct lu_fid *f1)
442 __diff_normalize(fid_seq(f0), fid_seq(f1)) ?:
443 __diff_normalize(fid_oid(f0), fid_oid(f1)) ?:
444 __diff_normalize(fid_ver(f0), fid_ver(f1));
449 /** \defgroup lu_dir lu_dir
452 * Layout of readdir pages, as transmitted on wire.
455 struct lu_fid lde_fid;
464 __u64 ldp_hash_start;
468 struct lu_dirent ldp_entries[0];
471 enum lu_dirpage_flags {
475 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
477 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
480 return dp->ldp_entries;
483 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
485 struct lu_dirent *next;
487 if (le16_to_cpu(ent->lde_reclen) != 0)
488 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
495 static inline int lu_dirent_size(struct lu_dirent *ent)
497 if (le16_to_cpu(ent->lde_reclen) == 0) {
498 return (sizeof(*ent) +
499 le16_to_cpu(ent->lde_namelen) + 7) & ~7;
501 return le16_to_cpu(ent->lde_reclen);
504 #define DIR_END_OFF 0xfffffffffffffffeULL
508 struct lustre_handle {
511 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
513 static inline int lustre_handle_is_used(struct lustre_handle *lh)
515 return lh->cookie != 0ull;
518 static inline int lustre_handle_equal(struct lustre_handle *lh1,
519 struct lustre_handle *lh2)
521 return lh1->cookie == lh2->cookie;
524 static inline void lustre_handle_copy(struct lustre_handle *tgt,
525 struct lustre_handle *src)
527 tgt->cookie = src->cookie;
530 /* flags for lm_flags */
531 #define MSGHDR_AT_SUPPORT 0x1
533 #define lustre_msg lustre_msg_v2
534 /* we depend on this structure to be 8-byte aligned */
535 /* this type is only endian-adjusted in lustre_unpack_msg() */
536 struct lustre_msg_v2 {
548 /* without gss, ptlrpc_body is put at the first buffer. */
549 #define PTLRPC_NUM_VERSIONS 4
551 struct lustre_handle pb_handle;
558 __u64 pb_last_committed;
563 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
564 __u32 pb_service_time; /* for rep, actual service time */
567 /* VBR: pre-versions */
568 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
569 /* padding for future needs */
573 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
575 /* message body offset for lustre_msg_v2 */
576 /* ptlrpc body offset in all request/reply messages */
577 #define MSG_PTLRPC_BODY_OFF 0
579 /* normal request/reply message record offset */
580 #define REQ_REC_OFF 1
581 #define REPLY_REC_OFF 1
583 /* ldlm request message body offset */
584 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
585 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
587 /* ldlm intent lock message body offset */
588 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
589 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
591 /* ldlm reply message body offset */
592 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
593 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
595 /* Flags that are operation-specific go in the top 16 bits. */
596 #define MSG_OP_FLAG_MASK 0xffff0000
597 #define MSG_OP_FLAG_SHIFT 16
599 /* Flags that apply to all requests are in the bottom 16 bits */
600 #define MSG_GEN_FLAG_MASK 0x0000ffff
601 #define MSG_LAST_REPLAY 0x0001
602 #define MSG_RESENT 0x0002
603 #define MSG_REPLAY 0x0004
604 /* #define MSG_AT_SUPPORT 0x0008
605 * This was used in early prototypes of adaptive timeouts, and while there
606 * shouldn't be any users of that code there also isn't a need for using this
607 * bits. Defer usage until at least 1.10 to avoid potential conflict. */
608 #define MSG_DELAY_REPLAY 0x0010
609 #define MSG_VERSION_REPLAY 0x0020
610 #define MSG_REQ_REPLAY_DONE 0x0040
611 #define MSG_LOCK_REPLAY_DONE 0x0080
614 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
617 #define MSG_CONNECT_RECOVERING 0x1
618 #define MSG_CONNECT_RECONNECT 0x2
619 #define MSG_CONNECT_REPLAYABLE 0x4
620 //#define MSG_CONNECT_PEER 0x8
621 #define MSG_CONNECT_LIBCLIENT 0x10
622 #define MSG_CONNECT_INITIAL 0x20
623 #define MSG_CONNECT_ASYNC 0x40
624 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
625 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
628 #define OBD_CONNECT_RDONLY 0x00000001ULL /* client allowed read-only access */
629 #define OBD_CONNECT_INDEX 0x00000002ULL /* connect to specific LOV idx */
630 #define OBD_CONNECT_GRANT 0x00000008ULL /* OSC acquires grant at connect */
631 #define OBD_CONNECT_SRVLOCK 0x00000010ULL /* server takes locks for client */
632 #define OBD_CONNECT_VERSION 0x00000020ULL /* Server supports versions in ocd */
633 #define OBD_CONNECT_REQPORTAL 0x00000040ULL /* Separate portal for non-IO reqs */
634 #define OBD_CONNECT_ACL 0x00000080ULL /* client uses access control lists */
635 #define OBD_CONNECT_XATTR 0x00000100ULL /* client using extended attributes*/
636 #define OBD_CONNECT_TRUNCLOCK 0x00000400ULL /* locks on server for punch b=9528 */
637 #define OBD_CONNECT_IBITS 0x00001000ULL /* support for inodebits locks */
638 #define OBD_CONNECT_JOIN 0x00002000ULL /* files can be concatenated */
639 #define OBD_CONNECT_ATTRFID 0x00004000ULL /* Server supports GetAttr By Fid */
640 #define OBD_CONNECT_NODEVOH 0x00008000ULL /* No open handle for special nodes */
641 #define OBD_CONNECT_LCL_CLIENT 0x00010000ULL /* local 1.8 client */
642 #define OBD_CONNECT_RMT_CLIENT 0x00020000ULL /* Remote 1.8 client */
643 #define OBD_CONNECT_BRW_SIZE 0x00040000ULL /* Max bytes per rpc */
644 #define OBD_CONNECT_QUOTA64 0x00080000ULL /* 64bit qunit_data.qd_count b=10707*/
645 #define OBD_CONNECT_MDS_CAPA 0x00100000ULL /* MDS capability */
646 #define OBD_CONNECT_OSS_CAPA 0x00200000ULL /* OSS capability */
647 #define OBD_CONNECT_CANCELSET 0x00400000ULL /* Early batched cancels. */
648 #define OBD_CONNECT_SOM 0x00800000ULL /* SOM feature */
649 #define OBD_CONNECT_AT 0x01000000ULL /* client uses adaptive timeouts */
650 #define OBD_CONNECT_LRU_RESIZE 0x02000000ULL /* Lru resize feature. */
651 #define OBD_CONNECT_MDS_MDS 0x04000000ULL /* MDS-MDS connection*/
652 #define OBD_CONNECT_REAL 0x08000000ULL /* real connection */
653 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /* shrink/enlarge qunit b=10600 */
654 #define OBD_CONNECT_CKSUM 0x20000000ULL /* support several cksum algos */
655 #define OBD_CONNECT_FID 0x40000000ULL /* FID is supported by server */
656 #define OBD_CONNECT_LOV_V3 0x100000000ULL /* client supports lov v3 ea */
658 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
659 * and lustre/utils/wirecheck.c */
661 #ifdef HAVE_LRU_RESIZE_SUPPORT
662 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
664 #define LRU_RESIZE_CONNECT_FLAG 0
667 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
668 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
669 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
670 OBD_CONNECT_NODEVOH |/* OBD_CONNECT_ATTRFID |*/\
671 OBD_CONNECT_LCL_CLIENT | \
672 OBD_CONNECT_RMT_CLIENT | \
673 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | \
674 OBD_CONNECT_MDS_MDS | OBD_CONNECT_CANCELSET | \
676 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_AT | \
678 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
679 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
680 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
681 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
682 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET | \
683 OBD_CONNECT_CKSUM | LRU_RESIZE_CONNECT_FLAG | \
685 #define ECHO_CONNECT_SUPPORTED (0)
686 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT)
688 #define MAX_QUOTA_COUNT32 (0xffffffffULL)
690 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
691 ((patch)<<8) + (fix))
692 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
693 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
694 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
695 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
697 /* This structure is used for both request and reply.
699 * If we eventually have separate connect data for different types, which we
700 * almost certainly will, then perhaps we stick a union in here. */
701 struct obd_connect_data {
702 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
703 __u32 ocd_version; /* lustre release version number */
704 __u32 ocd_grant; /* initial cache grant amount (bytes) */
705 __u32 ocd_index; /* LOV index to connect to */
706 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
707 __u64 ocd_ibits_known; /* inode bits this client understands */
708 __u32 ocd_nllu; /* non-local-lustre-user */
709 __u32 ocd_nllg; /* non-local-lustre-group */
710 __u64 ocd_transno; /* first transno from client to be replayed */
711 __u32 ocd_group; /* MDS group on OST */
712 __u32 ocd_cksum_types; /* supported checksum algorithms */
713 __u64 padding1; /* also fix lustre_swab_connect */
714 __u64 padding2; /* also fix lustre_swab_connect */
717 extern void lustre_swab_connect(struct obd_connect_data *ocd);
720 * Supported checksum algorithms. Up to 32 checksum types are supported.
721 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
722 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
723 * algorithm and also the OBD_FL_CKSUM* flags.
726 OBD_CKSUM_CRC32 = 0x00000001,
727 OBD_CKSUM_ADLER = 0x00000002,
731 * OST requests: OBDO & OBD request records
736 OST_REPLY = 0, /* reply ? */
756 #define OST_FIRST_OPC OST_REPLY
758 typedef __u64 obd_id;
759 typedef __u64 obd_gr;
760 typedef __u64 obd_time;
761 typedef __u64 obd_size;
762 typedef __u64 obd_off;
763 typedef __u64 obd_blocks;
764 typedef __u32 obd_blksize;
765 typedef __u32 obd_mode;
766 typedef __u32 obd_uid;
767 typedef __u32 obd_gid;
768 typedef __u32 obd_flag;
769 typedef __u64 obd_valid;
770 typedef __u32 obd_count;
772 #define OBD_FL_INLINEDATA (0x00000001)
773 #define OBD_FL_OBDMDEXISTS (0x00000002)
774 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
775 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
776 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
777 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
778 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
779 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
780 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
783 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
784 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
787 #define OBD_FL_TRUNCLOCK (0x00000800)
792 #define OBD_FL_CKSUM_CRC32 (0x00001000)
793 #define OBD_FL_CKSUM_ADLER (0x00002000)
794 #define OBD_FL_CKSUM_ALL (OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER)
796 #define LOV_MAGIC_V1 0x0BD10BD0
797 #define LOV_MAGIC LOV_MAGIC_V1
798 #define LOV_MAGIC_JOIN 0x0BD20BD0
799 #define LOV_MAGIC_V3 0x0BD30BD0
801 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
802 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
803 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
804 #define LOV_PATTERN_CMOBD 0x200
806 #define LOV_OBJECT_GROUP_DEFAULT ~0ULL
807 #define LOV_OBJECT_GROUP_CLEAR 0ULL
809 #define lov_ost_data lov_ost_data_v1
810 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
811 __u64 l_object_id; /* OST object ID */
812 __u64 l_object_gr; /* OST object group (creating MDS number) */
813 __u32 l_ost_gen; /* generation of this l_ost_idx */
814 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
817 #define lov_mds_md lov_mds_md_v1
818 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
819 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
820 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
821 __u64 lmm_object_id; /* LOV object ID */
822 __u64 lmm_object_gr; /* LOV object group */
823 __u32 lmm_stripe_size; /* size of stripe in bytes */
824 __u32 lmm_stripe_count; /* num stripes in use for this object */
825 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
828 /* extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm); */
830 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
831 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
833 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
834 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
835 #define XATTR_NAME_LOV "trusted.lov"
837 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
838 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
839 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
840 __u64 lmm_object_id; /* LOV object ID */
841 __u64 lmm_object_gr; /* LOV object group */
842 __u32 lmm_stripe_size; /* size of stripe in bytes */
843 __u32 lmm_stripe_count; /* num stripes in use for this object */
844 char lmm_pool_name[LOV_MAXPOOLNAME]; /* must be 32bit aligned */
845 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
849 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
850 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
851 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
852 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
853 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
854 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
855 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
856 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
857 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
858 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
859 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
860 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
861 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
862 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
863 /*#define OBD_MD_FLINLINE (0x00008000ULL) inline data. used until 1.6.5 */
864 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
865 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
866 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
867 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
868 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
869 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
870 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
871 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
872 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
873 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
874 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
875 /* ->mds if epoch opens or closes */
876 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
877 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
878 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
879 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
880 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
882 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
883 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
884 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
886 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
887 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
888 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
889 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
890 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
891 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
892 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
893 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
894 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
896 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
897 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
898 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
899 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
901 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
902 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
903 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
904 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
905 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
907 /* don't forget obdo_fid which is way down at the bottom so it can
908 * come after the definition of llog_cookie */
921 __u32 os_state; /* positive error code on server */
933 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
934 #define OBD_STATFS_NODELAY 0x0001 /* requests should be send without delay
935 * and resends for avoid deadlocks */
937 #define OBD_STATFS_FROM_CACHE 0x0002 /* the statfs callback should not update
940 /* ost_body.data values for OST_BRW */
942 #define OBD_BRW_READ 0x01
943 #define OBD_BRW_WRITE 0x02
944 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
945 #define OBD_BRW_SYNC 0x08
946 #define OBD_BRW_CHECK 0x10
947 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
948 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
949 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
950 #define OBD_BRW_NOQUOTA 0x100
951 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
953 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
955 #define OST_MIN_PRECREATE 32
956 #define OST_MAX_PRECREATE 20000
965 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
967 /* multiple of 8 bytes => can array */
968 struct niobuf_remote {
974 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
976 /* lock value block communicated between the filter and llite */
978 /* OST_LVB_ERR_INIT is needed because the return code in rc is
979 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
980 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
981 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
982 #define OST_LVB_IS_ERR(blocks) \
983 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
984 #define OST_LVB_SET_ERR(blocks, rc) \
985 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
986 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
996 extern void lustre_swab_ost_lvb(struct ost_lvb *);
1005 MDS_GETATTR_NAME = 34,
1010 MDS_DISCONNECT = 39,
1016 MDS_DONE_WRITING = 45,
1018 MDS_QUOTACHECK = 47,
1021 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1027 #define MDS_FIRST_OPC MDS_GETATTR
1044 } mds_reint_t, mdt_reint_t;
1046 extern void lustre_swab_generic_32s (__u32 *val);
1048 /* the disposition of the intent outlines what was executed */
1049 #define DISP_IT_EXECD 0x00000001
1050 #define DISP_LOOKUP_EXECD 0x00000002
1051 #define DISP_LOOKUP_NEG 0x00000004
1052 #define DISP_LOOKUP_POS 0x00000008
1053 #define DISP_OPEN_CREATE 0x00000010
1054 #define DISP_OPEN_OPEN 0x00000020
1055 #define DISP_ENQ_COMPLETE 0x00400000
1056 #define DISP_ENQ_OPEN_REF 0x00800000
1057 #define DISP_ENQ_CREATE_REF 0x01000000
1058 #define DISP_OPEN_LOCK 0x02000000
1060 /* INODE LOCK PARTS */
1061 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1062 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1063 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1065 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
1066 #define MDS_INODELOCK_MAXSHIFT 2
1067 /* This FULL lock is useful to take on unlink sort of operations */
1068 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1071 __u64 id; /* holds object id */
1072 __u32 generation; /* holds object generation */
1074 __u32 f_type; /* holds object type or stripe idx when passing it to
1075 * OST for saving into EA. */
1078 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1080 #define MDS_STATUS_CONN 1
1081 #define MDS_STATUS_LOV 2
1083 struct mds_status_req {
1088 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
1090 /* mdt_thread_info.mti_flags. */
1092 /* The flag indicates Size-on-MDS attributes are changed. */
1093 MF_SOM_CHANGE = (1 << 0),
1094 /* Flags indicates an epoch opens or closes. */
1095 MF_EPOCH_OPEN = (1 << 1),
1096 MF_EPOCH_CLOSE = (1 << 2),
1097 MF_MDC_CANCEL_FID1 = (1 << 3),
1098 MF_MDC_CANCEL_FID2 = (1 << 4),
1099 MF_MDC_CANCEL_FID3 = (1 << 5),
1100 MF_MDC_CANCEL_FID4 = (1 << 6),
1103 #define MF_SOM_LOCAL_FLAGS (MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID2 | \
1104 MF_MDC_CANCEL_FID3 | MF_MDC_CANCEL_FID4)
1106 #define MDS_BFLAG_UNCOMMITTED_WRITES 0x1
1107 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
1109 /* these should be identical to their EXT3_*_FL counterparts, and are
1110 * redefined here only to avoid dragging in ext3_fs.h */
1111 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
1112 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
1113 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
1114 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
1115 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1118 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
1119 * and we need to decode these into local S_* flags in the inode. Otherwise
1120 * we pass flags straight through (see bug 9486). */
1121 static inline int ll_ext_to_inode_flags(int flags)
1123 return (flags & MDS_BFLAG_EXT_FLAGS) ?
1124 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
1125 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
1126 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
1127 #if defined(S_DIRSYNC)
1128 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1130 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1131 (flags & ~MDS_BFLAG_EXT_FLAGS);
1134 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1135 * and we pass these straight through. Otherwise we need to convert from
1136 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1137 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1139 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1140 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1141 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1142 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1143 #if defined(S_DIRSYNC)
1144 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1146 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1153 struct lustre_handle handle;
1155 __u64 size; /* Offset, in the case of MDS_READPAGE */
1159 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1161 __u64 ino; /* for 1.6 compatibility */
1168 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1170 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1171 __u32 generation; /* for 1.6 compatibility */
1176 __u32 max_cookiesize;
1177 __u32 padding_4; /* also fix lustre_swab_mdt_body */
1183 struct lustre_handle handle;
1185 __u64 size; /* Offset, in the case of MDS_READPAGE */
1189 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1198 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1200 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1206 __u32 max_cookiesize;
1207 __u32 padding_4; /* also fix lustre_swab_mds_body */
1210 extern void lustre_swab_mds_body (struct mds_body *b);
1211 extern void lustre_swab_mdt_body (struct mdt_body *b);
1214 struct lustre_handle handle;
1220 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1222 #define Q_QUOTACHECK 0x800100
1223 #define Q_INITQUOTA 0x800101 /* init slave limits */
1224 #define Q_GETOINFO 0x800102 /* get obd quota info */
1225 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1227 #define Q_TYPESET(oqc, type) \
1228 ((oqc)->qc_type == type || (oqc)->qc_type == UGQUOTA)
1230 #define Q_GETOCMD(oqc) \
1231 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1233 struct obd_quotactl {
1238 struct obd_dqinfo qc_dqinfo;
1239 struct obd_dqblk qc_dqblk;
1242 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1244 /* inode access permission for remote user, the inode info are omitted,
1245 * for client knows them. */
1246 struct mds_remote_perm {
1251 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1254 /* permissions for md_perm.mp_perm */
1256 CFS_SETUID_PERM = 0x01,
1257 CFS_SETGID_PERM = 0x02,
1258 CFS_SETGRP_PERM = 0x04,
1259 CFS_RMTACL_PERM = 0x08
1262 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1264 struct mdt_remote_perm {
1269 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1272 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1274 struct mds_rec_setattr {
1281 struct ll_fid sa_fid;
1282 __u64 sa_valid; /* MDS_ATTR_* attributes */
1289 __u32 sa_attr_flags;
1290 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1293 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1295 struct mdt_rec_setattr {
1302 struct lu_fid sa_fid;
1311 __u32 sa_attr_flags;
1319 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1322 * Attribute flags used in mds_rec_setattr::sa_valid.
1323 * The kernel's #defines for ATTR_* should not be used over the network
1324 * since the client and MDS may run different kernels (see bug 13828)
1325 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1327 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1328 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1329 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1330 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1331 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1332 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1333 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1334 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1335 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1336 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1337 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1338 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1339 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1340 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1341 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1342 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1345 #define FMODE_READ 00000001
1346 #define FMODE_WRITE 00000002
1349 #define FMODE_EPOCH 01000000
1350 #define FMODE_EPOCHLCK 02000000
1351 #define FMODE_SOM 04000000
1352 #define FMODE_CLOSED 0
1354 #define MDS_OPEN_CREATED 00000010
1355 #define MDS_OPEN_CROSS 00000020
1357 #define MDS_FMODE_EXEC 00000004
1358 #define MDS_OPEN_CREAT 00000100
1359 #define MDS_OPEN_EXCL 00000200
1360 #define MDS_OPEN_TRUNC 00001000
1361 #define MDS_OPEN_APPEND 00002000
1362 #define MDS_OPEN_SYNC 00010000
1363 #define MDS_OPEN_DIRECTORY 00200000
1365 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1366 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1367 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1368 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1369 * with default ACL */
1370 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1371 * actually, this is for create, not
1372 * conflict with other open flags */
1373 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1374 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1375 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1377 /* permission for create non-directory file */
1378 #define MAY_CREATE (1 << 7)
1379 /* permission for create directory file */
1380 #define MAY_LINK (1 << 8)
1381 /* permission for delete from the directory */
1382 #define MAY_UNLINK (1 << 9)
1383 /* source's permission for rename */
1384 #define MAY_RENAME_SRC (1 << 10)
1385 /* target's permission for rename */
1386 #define MAY_RENAME_TAR (1 << 11)
1387 /* part (parent's) VTX permission check */
1388 #define MAY_VTX_PART (1 << 12)
1389 /* full VTX permission check */
1390 #define MAY_VTX_FULL (1 << 13)
1391 /* lfs rgetfacl permission check */
1392 #define MAY_RGETFACL (1 << 14)
1395 MDS_CHECK_SPLIT = 1 << 0,
1396 MDS_CROSS_REF = 1 << 1,
1397 MDS_VTX_BYPASS = 1 << 2,
1398 MDS_PERM_BYPASS = 1 << 3,
1402 struct mds_rec_join {
1403 struct ll_fid jr_fid;
1407 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1409 struct mdt_rec_join {
1410 struct lu_fid jr_fid;
1414 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1416 struct mds_rec_create {
1421 __u32 cr_flags; /* for use with open */
1423 struct ll_fid cr_fid;
1424 struct ll_fid cr_replayfid;
1428 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1429 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1430 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1431 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1432 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1435 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1437 struct mdt_rec_create {
1444 struct lu_fid cr_fid1;
1445 struct lu_fid cr_fid2;
1446 struct lustre_handle cr_old_handle; /* u64 handle in case of open replay */
1450 __u64 cr_padding_1; /* pad for 64 bits*/
1453 __u32 cr_flags; /* for use with open */
1459 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1461 struct mds_rec_link {
1468 struct ll_fid lk_fid1;
1469 struct ll_fid lk_fid2;
1471 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1472 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1473 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1474 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1477 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1479 struct mdt_rec_link {
1486 struct lu_fid lk_fid1;
1487 struct lu_fid lk_fid2;
1501 struct mds_rec_unlink {
1508 struct ll_fid ul_fid1;
1509 struct ll_fid ul_fid2;
1511 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1512 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1513 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1514 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1517 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1519 struct mdt_rec_unlink {
1526 struct lu_fid ul_fid1;
1527 struct lu_fid ul_fid2;
1541 struct mds_rec_rename {
1548 struct ll_fid rn_fid1;
1549 struct ll_fid rn_fid2;
1551 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1552 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1553 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1554 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1557 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1559 struct mdt_rec_rename {
1566 struct lu_fid rn_fid1;
1567 struct lu_fid rn_fid2;
1573 __u32 rn_bias; /* some operation flags */
1574 __u32 rn_mode; /* cross-ref rename has mode */
1581 struct mdt_rec_setxattr {
1588 struct lu_fid sx_fid;
1589 __u64 sx_padding_1; /* These three members are lu_fid size */
1601 __u32 sx_padding_10;
1602 __u32 sx_padding_11;
1605 struct mdt_rec_reint {
1612 struct lu_fid rr_fid1;
1613 struct lu_fid rr_fid2;
1621 __u32 rr_padding_1; /* also fix lustre_swab_mdt_rec_reint */
1622 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
1623 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
1624 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1627 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
1630 __u32 ld_tgt_count; /* how many MDS's */
1631 __u32 ld_active_tgt_count; /* how many active */
1632 struct obd_uuid ld_uuid;
1635 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1642 extern void lustre_swab_md_fld (struct md_fld *mf);
1647 FLD_FIRST_OPC = FLD_QUERY
1653 SEQ_FIRST_OPC = SEQ_QUERY
1657 SEQ_ALLOC_SUPER = 0,
1662 * LOV data structures
1665 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1666 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1667 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
1669 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1670 /* The size of the buffer the lov/mdc reserves for the
1671 * array of UUIDs returned by the MDS. With the current
1672 * protocol, this will limit the max number of OSTs per LOV */
1674 #define LOV_DESC_MAGIC 0xB0CCDE5C
1676 /* LOV settings descriptor (should only contain static info) */
1678 __u32 ld_tgt_count; /* how many OBD's */
1679 __u32 ld_active_tgt_count; /* how many active */
1680 __u32 ld_default_stripe_count; /* how many objects are used */
1681 __u32 ld_pattern; /* default PATTERN_RAID0 */
1682 __u64 ld_default_stripe_size; /* in bytes */
1683 __u64 ld_default_stripe_offset; /* in bytes */
1684 __u32 ld_padding_0; /* unused */
1685 __u32 ld_qos_maxage; /* in second */
1686 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1687 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1688 struct obd_uuid ld_uuid;
1691 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1693 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1698 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1703 LDLM_BL_CALLBACK = 104,
1704 LDLM_CP_CALLBACK = 105,
1705 LDLM_GL_CALLBACK = 106,
1708 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1710 #define RES_NAME_SIZE 4
1711 struct ldlm_res_id {
1712 __u64 name[RES_NAME_SIZE];
1715 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1731 #define LCK_MODE_NUM 8
1741 #define LDLM_MIN_TYPE LDLM_PLAIN
1743 struct ldlm_extent {
1749 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
1750 struct ldlm_extent *ex2)
1752 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
1755 struct ldlm_inodebits {
1762 __u64 blocking_export; /* not actually used over the wire */
1763 __u32 blocking_pid; /* not actually used over the wire */
1767 /* it's important that the fields of the ldlm_extent structure match
1768 * the first fields of the ldlm_flock structure because there is only
1769 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1770 * this ever changes we will need to swab the union differently based
1771 * on the resource type. */
1774 struct ldlm_extent l_extent;
1775 struct ldlm_flock l_flock;
1776 struct ldlm_inodebits l_inodebits;
1777 } ldlm_policy_data_t;
1779 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1781 struct ldlm_intent {
1785 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1787 struct ldlm_resource_desc {
1788 ldlm_type_t lr_type;
1789 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1790 struct ldlm_res_id lr_name;
1793 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1795 struct ldlm_lock_desc {
1796 struct ldlm_resource_desc l_resource;
1797 ldlm_mode_t l_req_mode;
1798 ldlm_mode_t l_granted_mode;
1799 ldlm_policy_data_t l_policy_data;
1802 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1804 #define LDLM_LOCKREQ_HANDLES 2
1805 #define LDLM_ENQUEUE_CANCEL_OFF 1
1807 struct ldlm_request {
1810 struct ldlm_lock_desc lock_desc;
1811 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
1814 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1816 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
1817 * Otherwise, 2 are available. */
1818 #define ldlm_request_bufsize(count,type) \
1820 int _avail = LDLM_LOCKREQ_HANDLES; \
1821 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
1822 sizeof(struct ldlm_request) + \
1823 (count > _avail ? count - _avail : 0) * \
1824 sizeof(struct lustre_handle); \
1829 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1830 struct ldlm_lock_desc lock_desc;
1831 struct lustre_handle lock_handle;
1832 __u64 lock_policy_res1;
1833 __u64 lock_policy_res2;
1836 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1839 * Opcodes for mountconf (mgs and mgc)
1844 MGS_EXCEPTION, /* node died, etc. */
1845 MGS_TARGET_REG, /* whenever target starts up */
1850 #define MGS_FIRST_OPC MGS_CONNECT
1852 #define MGS_PARAM_MAXLEN 1024
1853 #define KEY_SET_INFO "set_info"
1855 struct mgs_send_param {
1856 char mgs_param[MGS_PARAM_MAXLEN];
1859 /* We pass this info to the MGS so it can write config logs */
1860 #define MTI_NAME_MAXLEN 64
1861 #define MTI_PARAM_MAXLEN 4096
1862 #define MTI_NIDS_MAX 32
1863 struct mgs_target_info {
1864 __u32 mti_lustre_ver;
1865 __u32 mti_stripe_index;
1866 __u32 mti_config_ver;
1868 __u32 mti_nid_count;
1869 __u32 padding; /* 64 bit align */
1870 char mti_fsname[MTI_NAME_MAXLEN];
1871 char mti_svname[MTI_NAME_MAXLEN];
1872 char mti_uuid[sizeof(struct obd_uuid)];
1873 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
1874 char mti_params[MTI_PARAM_MAXLEN];
1877 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
1879 /* Config marker flags (in config log) */
1880 #define CM_START 0x01
1882 #define CM_SKIP 0x04
1883 #define CM_UPGRADE146 0x08
1884 #define CM_EXCLUDE 0x10
1885 #define CM_START_SKIP (CM_START | CM_SKIP)
1888 __u32 cm_step; /* aka config version */
1890 __u32 cm_vers; /* lustre release version number */
1891 __u32 padding; /* 64 bit align */
1892 __u64 cm_createtime; /*when this record was first created */
1893 __u64 cm_canceltime; /*when this record is no longer valid*/
1894 char cm_tgtname[MTI_NAME_MAXLEN];
1895 char cm_comment[MTI_NAME_MAXLEN];
1898 extern void lustre_swab_cfg_marker(struct cfg_marker *marker,
1899 int swab, int size);
1902 * Opcodes for multiple servers.
1911 #define OBD_FIRST_OPC OBD_PING
1913 /* catalog of log objects */
1915 /** Identifier for a single log object */
1920 } __attribute__((packed));
1922 /** Records written to the CATALOGS list */
1923 #define CATLIST "CATALOGS"
1925 struct llog_logid lci_logid;
1929 } __attribute__((packed));
1931 /** join file lov mds md*/
1932 struct lov_mds_md_join {
1933 struct lov_mds_md lmmj_md;
1934 /*join private info*/
1935 struct llog_logid lmmj_array_id; /*array object id*/
1936 __u32 lmmj_extent_count; /*array extent count*/
1939 /* Log data record types - there is no specific reason that these need to
1940 * be related to the RPC opcodes, but no reason not to (may be handy later?)
1942 #define LLOG_OP_MAGIC 0x10600000
1943 #define LLOG_OP_MASK 0xfff00000
1946 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
1947 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
1948 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
1949 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
1950 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
1951 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
1952 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
1953 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
1954 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
1955 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
1956 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
1960 * for now, continue to support old pad records which have 0 for their
1961 * type but still need to be swabbed for their length
1963 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
1964 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
1965 __swab32(LLOG_OP_MAGIC) || \
1966 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
1968 /** Log record header - stored in little endian order.
1969 * Each record must start with this struct, end with a llog_rec_tail,
1970 * and be a multiple of 256 bits in size.
1972 struct llog_rec_hdr {
1979 struct llog_rec_tail {
1984 struct llog_logid_rec {
1985 struct llog_rec_hdr lid_hdr;
1986 struct llog_logid lid_id;
1992 struct llog_rec_tail lid_tail;
1993 } __attribute__((packed));
1995 /** MDS extent description
1996 * It is for joined file extent info, each extent info for joined file
1997 * just like (start, end, lmm).
1999 struct mds_extent_desc {
2000 __u64 med_start; /* extent start */
2001 __u64 med_len; /* extent length */
2002 struct lov_mds_md med_lmm; /* extent's lmm */
2005 /** Joined file array extent log record*/
2006 struct llog_array_rec {
2007 struct llog_rec_hdr lmr_hdr;
2008 struct mds_extent_desc lmr_med;
2009 struct llog_rec_tail lmr_tail;
2012 struct llog_create_rec {
2013 struct llog_rec_hdr lcr_hdr;
2014 struct ll_fid lcr_fid;
2018 struct llog_rec_tail lcr_tail;
2019 } __attribute__((packed));
2021 struct llog_orphan_rec {
2022 struct llog_rec_hdr lor_hdr;
2026 struct llog_rec_tail lor_tail;
2027 } __attribute__((packed));
2029 struct llog_unlink_rec {
2030 struct llog_rec_hdr lur_hdr;
2034 struct llog_rec_tail lur_tail;
2035 } __attribute__((packed));
2037 struct llog_setattr_rec {
2038 struct llog_rec_hdr lsr_hdr;
2044 struct llog_rec_tail lsr_tail;
2045 } __attribute__((packed));
2047 struct llog_size_change_rec {
2048 struct llog_rec_hdr lsc_hdr;
2049 struct ll_fid lsc_fid;
2052 struct llog_rec_tail lsc_tail;
2053 } __attribute__((packed));
2058 } __attribute__((packed));
2060 struct llog_gen_rec {
2061 struct llog_rec_hdr lgr_hdr;
2062 struct llog_gen lgr_gen;
2063 struct llog_rec_tail lgr_tail;
2065 /* On-disk header structure of each log object, stored in little endian order */
2066 #define LLOG_CHUNK_SIZE 8192
2067 #define LLOG_HEADER_SIZE (96)
2068 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2070 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2072 /* flags for the logs */
2073 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
2074 #define LLOG_F_IS_CAT 0x2
2075 #define LLOG_F_IS_PLAIN 0x4
2077 struct llog_log_hdr {
2078 struct llog_rec_hdr llh_hdr;
2079 __u64 llh_timestamp;
2081 __u32 llh_bitmap_offset;
2085 /* for a catalog the first plain slot is next to it */
2086 struct obd_uuid llh_tgtuuid;
2087 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2088 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2089 struct llog_rec_tail llh_tail;
2090 } __attribute__((packed));
2092 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
2093 llh->llh_bitmap_offset - \
2094 sizeof(llh->llh_tail)) * 8)
2096 /** log cookies are used to reference a specific log file and a record therein */
2097 struct llog_cookie {
2098 struct llog_logid lgc_lgl;
2102 } __attribute__((packed));
2104 /** llog protocol */
2105 enum llogd_rpc_ops {
2106 LLOG_ORIGIN_HANDLE_CREATE = 501,
2107 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2108 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2109 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2110 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2111 LLOG_ORIGIN_CONNECT = 506,
2112 LLOG_CATINFO = 507, /* for lfs catinfo */
2113 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2114 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2116 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2120 struct llog_logid lgd_logid;
2122 __u32 lgd_llh_flags;
2124 __u32 lgd_saved_index;
2126 __u64 lgd_cur_offset;
2127 } __attribute__((packed));
2129 struct llogd_conn_body {
2130 struct llog_gen lgdc_gen;
2131 struct llog_logid lgdc_logid;
2132 __u32 lgdc_ctxt_idx;
2133 } __attribute__((packed));
2135 struct lov_user_ost_data_join { /* per-stripe data structure */
2136 __u64 l_extent_start; /* extent start*/
2137 __u64 l_extent_end; /* extent end*/
2138 __u64 l_object_id; /* OST object ID */
2139 __u64 l_object_gr; /* OST object group (creating MDS number) */
2140 __u32 l_ost_gen; /* generation of this OST index */
2141 __u32 l_ost_idx; /* OST index in LOV */
2142 } __attribute__((packed));
2144 struct lov_user_md_join { /* LOV EA user data (host-endian) */
2145 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
2146 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
2147 __u64 lmm_object_id; /* LOV object ID */
2148 __u64 lmm_object_gr; /* LOV object group */
2149 __u32 lmm_stripe_size; /* size of stripe in bytes */
2150 __u32 lmm_stripe_count; /* num stripes in use for this object */
2151 __u32 lmm_extent_count; /* extent count of lmm*/
2152 __u64 lmm_tree_id; /* mds tree object id */
2153 __u64 lmm_tree_gen; /* mds tree object gen */
2154 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
2155 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
2156 } __attribute__((packed));
2158 /* Note: 64-bit types are 64-bit aligned in structure */
2160 obd_valid o_valid; /* hot fields in this obdo */
2164 obd_size o_size; /* o_size-o_blocks == ost_lvb */
2168 obd_blocks o_blocks; /* brw: cli sent cached bytes */
2171 /* 32-bit fields start here: keep an even number of them via padding */
2172 obd_blksize o_blksize; /* optimal IO blocksize */
2173 obd_mode o_mode; /* brw: cli sent cache remain */
2177 obd_count o_nlink; /* brw: checksum */
2178 obd_count o_generation;
2179 obd_count o_misc; /* brw: o_dropped */
2180 __u32 o_easize; /* epoch in ost writes */
2182 __u32 o_stripe_idx; /* holds stripe idx */
2184 struct lustre_handle o_handle; /* brw: lock handle to prolong locks */
2185 struct llog_cookie o_lcookie; /* destroy: unlink cookie from MDS */
2194 #define o_dirty o_blocks
2195 #define o_undirty o_mode
2196 #define o_dropped o_misc
2197 #define o_cksum o_nlink
2199 extern void lustre_swab_obdo (struct obdo *o);
2201 /* request structure for OST's */
2207 /* Key for FIEMAP to be used in get_info calls */
2208 struct ll_fiemap_info_key {
2211 struct ll_user_fiemap fiemap;
2214 extern void lustre_swab_ost_body (struct ost_body *b);
2215 extern void lustre_swab_ost_last_id(obd_id *id);
2216 extern void lustre_swab_fiemap(struct ll_user_fiemap *fiemap);
2218 extern void lustre_swab_lov_user_md_v1(struct lov_user_md_v1 *lum);
2219 extern void lustre_swab_lov_user_md_v3(struct lov_user_md_v3 *lum);
2220 extern void lustre_swab_lov_user_md_objects(struct lov_user_ost_data *lod,
2222 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
2225 extern void lustre_swab_llogd_body (struct llogd_body *d);
2226 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
2227 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
2228 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
2229 struct llog_rec_tail *tail);
2232 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
2234 /* quota. fixed by tianzy for bug10707 */
2235 #define QUOTA_IS_GRP 0X1UL /* 0 is user, 1 is group. Used by qd_flags*/
2236 #define QUOTA_IS_BLOCK 0x2UL /* 0 is inode, 1 is block. Used by qd_flags*/
2239 __u32 qd_id; /* ID appiles to (uid, gid) */
2240 __u32 qd_flags; /* Quota type (USRQUOTA, GRPQUOTA) occupy one bit;
2241 * Block quota or file quota occupy one bit */
2242 __u64 qd_count; /* acquire/release count (bytes for block quota) */
2245 struct qunit_data_old {
2246 __u32 qd_id; /* ID appiles to (uid, gid) */
2247 __u32 qd_type; /* Quota type (USRQUOTA, GRPQUOTA) */
2248 __u32 qd_count; /* acquire/release count (bytes for block quota) */
2249 __u32 qd_isblk; /* Block quota or file quota */
2252 extern void lustre_swab_qdata(struct qunit_data *d);
2253 extern void lustre_swab_qdata_old(struct qunit_data_old *d);
2254 extern struct qunit_data *lustre_quota_old_to_new(struct qunit_data_old *d);
2255 extern struct qunit_data_old *lustre_quota_new_to_old(struct qunit_data *d);
2262 #define JOIN_FILE_ALIGN 4096
2264 /** security opcodes */
2267 SEC_CTX_INIT_CONT = 802,
2270 SEC_FIRST_OPC = SEC_CTX_INIT
2274 * capa related definitions
2276 #define CAPA_HMAC_MAX_LEN 64
2277 #define CAPA_HMAC_KEY_MAX_LEN 56
2279 /* NB take care when changing the sequence of elements this struct,
2280 * because the offset info is used in find_capa() */
2281 struct lustre_capa {
2282 struct lu_fid lc_fid; /* fid */
2283 __u64 lc_opc; /* operations allowed */
2284 __u32 lc_uid; /* uid, it is obsolete, but maybe used in
2285 * future, reserve it for 64-bits aligned.*/
2286 __u32 lc_flags; /* HMAC algorithm & flags */
2287 __u32 lc_keyid; /* key used for the capability */
2288 __u32 lc_timeout; /* capa timeout value (sec) */
2289 __u64 lc_expiry; /* expiry time (sec) */
2290 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /* HMAC */
2291 } __attribute__((packed));
2293 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
2295 /** lustre_capa::lc_opc */
2297 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
2298 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
2299 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
2300 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
2301 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
2302 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
2303 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
2304 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
2305 CAPA_OPC_META_WRITE = 1<<8, /**< write object meta data */
2306 CAPA_OPC_META_READ = 1<<9, /**< read object meta data */
2310 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2311 #define CAPA_OPC_MDS_ONLY \
2312 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2313 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2314 #define CAPA_OPC_OSS_ONLY \
2315 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC)
2316 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2317 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2319 /* MDS capability covers object capability for operations of body r/w
2320 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
2321 * while OSS capability only covers object capability for operations of
2322 * oss data(file content) r/w/truncate.
2324 static inline int capa_for_mds(struct lustre_capa *c)
2326 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2329 static inline int capa_for_oss(struct lustre_capa *c)
2331 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2334 /* lustre_capa::lc_hmac_alg */
2336 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
2340 #define CAPA_FL_MASK 0x00ffffff
2341 #define CAPA_HMAC_ALG_MASK 0xff000000
2343 struct lustre_capa_key {
2344 __u64 lk_mdsid; /**< mds# */
2345 __u32 lk_keyid; /**< key# */
2347 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
2348 } __attribute__((packed));
2350 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2352 /* quota check function */
2353 #define QUOTA_RET_OK 0 /**< return successfully */
2354 #define QUOTA_RET_NOQUOTA 1 /**< not support quota */
2355 #define QUOTA_RET_NOLIMIT 2 /**< quota limit isn't set */
2356 #define QUOTA_RET_ACQUOTA 3 /**< need to acquire extra quota */