1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * This file is part of Lustre, http://www.lustre.org
6 * Lustre wire protocol definitions.
9 * We assume all nodes are either little-endian or big-endian, and we
10 * always send messages in the sender's native format. The receiver
11 * detects the message format by checking the 'magic' field of the message
12 * (see lustre_msg_swabbed() below).
13 * ALL structs passing over the wire should be declared here. Structs
14 * that are used in interfaces with userspace should go in lustre_user.h.
16 * All structs being declared here should be built from simple fixed-size
17 * types (__u8, __u16, __u32, __u64) or be built from other types or
18 * structs also declared in this file. Similarly, all flags and magic
19 * values in those structs should also be declared here. This ensures
20 * that the Lustre wire protocol is not influenced by external dependencies.
22 * The only other acceptable items in this file are VERY SIMPLE accessor
23 * functions to avoid callers grubbing inside the structures, and the
24 * prototypes of the swabber functions for each struct. Nothing that
25 * depends on external functions or definitions should be in here.
27 * Structs must be properly aligned to put 64-bit values on an 8-byte
28 * boundary. Any structs being added here must also be added to
29 * utils/wirecheck.c and "make newwiretest" run to regenerate the
30 * utils/wiretest.c sources. This allows us to verify that wire structs
31 * have the proper alignment/size on all architectures.
33 * DO NOT CHANGE any of the structs, flags, values declared here and used
34 * in released Lustre versions. Some structs may have padding fields that
35 * can be used. Some structs might allow addition at the end (verify this
36 * in the code to ensure that new/old clients that see this larger struct
37 * do not fail, otherwise you need to implement protocol compatibility).
39 * Each wire type has corresponding 'lustre_swab_xxxtypexxx()' routines,
40 * implemented either here, inline (trivial implementations) or in
41 * ptlrpc/pack_generic.c. These 'swabbers' convert the type from "other"
42 * endian, in-place in the message buffer.
44 * A swabber takes a single pointer argument. The caller must already have
45 * verified that the length of the message buffer >= sizeof (type).
47 * For variable length types, a second 'lustre_swab_v_xxxtypexxx()' routine
48 * may be defined that swabs just the variable part, after the caller has
49 * verified that the message buffer is large enough.
52 #ifndef _LUSTRE_IDL_H_
53 #define _LUSTRE_IDL_H_
55 #include <libcfs/libcfs.h>
57 #include <lustre/types.h>
59 /* Defn's shared with user-space. */
60 #include <lustre/lustre_user.h>
65 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
66 * FOO_REPLY_PORTAL is for incoming replies on the FOO
67 * FOO_BULK_PORTAL is for incoming bulk on the FOO
70 #define CONNMGR_REQUEST_PORTAL 1
71 #define CONNMGR_REPLY_PORTAL 2
72 //#define OSC_REQUEST_PORTAL 3
73 #define OSC_REPLY_PORTAL 4
74 //#define OSC_BULK_PORTAL 5
75 #define OST_IO_PORTAL 6
76 #define OST_CREATE_PORTAL 7
77 #define OST_BULK_PORTAL 8
78 //#define MDC_REQUEST_PORTAL 9
79 #define MDC_REPLY_PORTAL 10
80 //#define MDC_BULK_PORTAL 11
81 #define MDS_REQUEST_PORTAL 12
82 //#define MDS_REPLY_PORTAL 13
83 #define MDS_BULK_PORTAL 14
84 #define LDLM_CB_REQUEST_PORTAL 15
85 #define LDLM_CB_REPLY_PORTAL 16
86 #define LDLM_CANCEL_REQUEST_PORTAL 17
87 #define LDLM_CANCEL_REPLY_PORTAL 18
88 //#define PTLBD_REQUEST_PORTAL 19
89 //#define PTLBD_REPLY_PORTAL 20
90 //#define PTLBD_BULK_PORTAL 21
91 #define MDS_SETATTR_PORTAL 22
92 #define MDS_READPAGE_PORTAL 23
93 #define MDS_MDS_PORTAL 24
95 #define MGC_REPLY_PORTAL 25
96 #define MGS_REQUEST_PORTAL 26
97 #define MGS_REPLY_PORTAL 27
98 #define OST_REQUEST_PORTAL 28
99 #define FLD_REQUEST_PORTAL 29
100 #define SEQ_METADATA_PORTAL 30
101 #define SEQ_DATA_PORTAL 31
102 #define SEQ_CONTROLLER_PORTAL 32
107 #define SVC_RUNNING 8
108 #define SVC_STOPPING 16
109 #define SVC_STOPPED 32
112 #define PTL_RPC_MSG_REQUEST 4711
113 #define PTL_RPC_MSG_ERR 4712
114 #define PTL_RPC_MSG_REPLY 4713
116 /* DON'T use swabbed values of MAGIC as magic! */
117 #define LUSTRE_MSG_MAGIC_V1 0x0BD00BD0
118 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
120 #define LUSTRE_MSG_MAGIC_V1_SWABBED 0xD00BD00B
121 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
123 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
125 #define PTLRPC_MSG_VERSION 0x00000003
126 #define LUSTRE_VERSION_MASK 0xffff0000
127 #define LUSTRE_OBD_VERSION 0x00010000
128 #define LUSTRE_MDS_VERSION 0x00020000
129 #define LUSTRE_OST_VERSION 0x00030000
130 #define LUSTRE_DLM_VERSION 0x00040000
131 #define LUSTRE_LOG_VERSION 0x00050000
132 #define LUSTRE_MGS_VERSION 0x00060000
134 typedef __u64 mdsno_t;
135 typedef __u64 seqno_t;
142 static inline __u64 range_space(struct lu_range *r)
144 return r->lr_end - r->lr_start;
147 static inline void range_zero(struct lu_range *r)
149 r->lr_start = r->lr_end = 0;
152 static inline int range_within(struct lu_range *r,
155 return s >= r->lr_start && s < r->lr_end;
158 static inline void range_alloc(struct lu_range *r,
162 r->lr_start = s->lr_start;
163 r->lr_end = s->lr_start + w;
167 static inline int range_is_sane(struct lu_range *r)
169 return (r->lr_end >= r->lr_start);
172 static inline int range_is_zero(struct lu_range *r)
174 return (r->lr_start == 0 && r->lr_end == 0);
177 static inline int range_is_exhausted(struct lu_range *r)
179 return range_space(r) == 0;
182 #define DRANGE "[%#16.16"LPF64"x-%#16.16"LPF64"x]"
184 #define PRANGE(range) \
189 __u64 f_seq; /* holds fid sequence. Lustre should support 2 ^ 64
190 * objects, thus even if one sequence has one object we
191 * will never reach this value. */
192 __u32 f_oid; /* fid number within its sequence. */
193 __u32 f_ver; /* holds fid version. */
200 /* initial fid id value */
201 LUSTRE_FID_INIT_OID = 1UL
204 /* get object sequence */
205 static inline __u64 fid_seq(const struct lu_fid *fid)
211 static inline __u32 fid_oid(const struct lu_fid *fid)
216 /* get object version */
217 static inline __u32 fid_ver(const struct lu_fid *fid)
222 static inline void fid_zero(struct lu_fid *fid)
224 memset(fid, 0, sizeof(*fid));
227 /* Normal FID sequence starts from this value, i.e. 1<<33 */
228 #define FID_SEQ_START 0x200000000ULL
230 /* IDIF sequence starts from this value, i.e. 1<<32 */
231 #define IDIF_SEQ_START 0x100000000ULL
234 * Check if a fid is igif or not.
235 * \param fid the fid to be tested.
236 * \return true if the fid is a igif; otherwise false.
238 static inline int fid_is_igif(const struct lu_fid *fid)
240 return fid_seq(fid) > 0 && fid_seq(fid) < IDIF_SEQ_START;
244 * Check if a fid is idif or not.
245 * \param fid the fid to be tested.
246 * \return true if the fid is a idif; otherwise false.
248 static inline int fid_is_idif(const struct lu_fid *fid)
250 return fid_seq(fid) >= IDIF_SEQ_START && fid_seq(fid) < FID_SEQ_START;
254 * Get inode number from a igif.
255 * \param fid a igif to get inode number from.
256 * \return inode number for the igif.
258 static inline ino_t lu_igif_ino(const struct lu_fid *fid)
264 * Get inode generation from a igif.
265 * \param fid a igif to get inode generation from.
266 * \return inode generation for the igif.
268 static inline __u32 lu_igif_gen(const struct lu_fid *fid)
273 #define DFID "[0x%16.16"LPF64"x/0x%8.8x:0x%8.8x]"
280 static inline void fid_cpu_to_le(struct lu_fid *dst, const struct lu_fid *src)
282 /* check that all fields are converted */
283 CLASSERT(sizeof *src ==
284 sizeof fid_seq(src) +
285 sizeof fid_oid(src) + sizeof fid_ver(src));
286 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
287 dst->f_seq = cpu_to_le64(fid_seq(src));
288 dst->f_oid = cpu_to_le32(fid_oid(src));
289 dst->f_ver = cpu_to_le32(fid_ver(src));
292 static inline void fid_le_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
294 /* check that all fields are converted */
295 CLASSERT(sizeof *src ==
296 sizeof fid_seq(src) +
297 sizeof fid_oid(src) + sizeof fid_ver(src));
298 dst->f_seq = le64_to_cpu(fid_seq(src));
299 dst->f_oid = le32_to_cpu(fid_oid(src));
300 dst->f_ver = le32_to_cpu(fid_ver(src));
301 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
304 static inline void fid_cpu_to_be(struct lu_fid *dst, const struct lu_fid *src)
306 /* check that all fields are converted */
307 CLASSERT(sizeof *src ==
308 sizeof fid_seq(src) +
309 sizeof fid_oid(src) + sizeof fid_ver(src));
310 LASSERTF(fid_is_igif(src) || fid_ver(src) == 0, DFID"\n", PFID(src));
311 dst->f_seq = cpu_to_be64(fid_seq(src));
312 dst->f_oid = cpu_to_be32(fid_oid(src));
313 dst->f_ver = cpu_to_be32(fid_ver(src));
316 static inline void fid_be_to_cpu(struct lu_fid *dst, const struct lu_fid *src)
318 /* check that all fields are converted */
319 CLASSERT(sizeof *src ==
320 sizeof fid_seq(src) +
321 sizeof fid_oid(src) + sizeof fid_ver(src));
322 dst->f_seq = be64_to_cpu(fid_seq(src));
323 dst->f_oid = be32_to_cpu(fid_oid(src));
324 dst->f_ver = be32_to_cpu(fid_ver(src));
325 LASSERTF(fid_is_igif(dst) || fid_ver(dst) == 0, DFID"\n", PFID(dst));
330 * Storage representation for fids.
332 * Variable size, first byte contains the length of the whole record.
337 char fp_area[sizeof(struct lu_fid)];
340 void fid_pack(struct lu_fid_pack *pack, const struct lu_fid *fid,
341 struct lu_fid *befider);
342 int fid_unpack(const struct lu_fid_pack *pack, struct lu_fid *fid);
347 static inline int fid_is_sane(const struct lu_fid *fid)
351 ((fid_seq(fid) >= FID_SEQ_START && fid_oid(fid) != 0
352 && fid_ver(fid) == 0) ||
356 static inline int fid_is_zero(const struct lu_fid *fid)
358 return fid_seq(fid) == 0 && fid_oid(fid) == 0;
361 extern void lustre_swab_lu_fid(struct lu_fid *fid);
362 extern void lustre_swab_lu_range(struct lu_range *range);
364 static inline int lu_fid_eq(const struct lu_fid *f0,
365 const struct lu_fid *f1)
367 /* Check that there is no alignment padding. */
368 CLASSERT(sizeof *f0 ==
369 sizeof f0->f_seq + sizeof f0->f_oid + sizeof f0->f_ver);
370 LASSERTF(fid_is_igif(f0) || fid_ver(f0) == 0, DFID, PFID(f0));
371 LASSERTF(fid_is_igif(f1) || fid_ver(f1) == 0, DFID, PFID(f1));
372 return memcmp(f0, f1, sizeof *f0) == 0;
376 * Layout of readdir pages, as transmitted on wire.
379 struct lu_fid lde_fid;
388 __u64 ldp_hash_start;
392 struct lu_dirent ldp_entries[0];
395 enum lu_dirpage_flags {
399 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
401 if (le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
404 return dp->ldp_entries;
407 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
409 struct lu_dirent *next;
411 if (le16_to_cpu(ent->lde_reclen) != 0)
412 next = ((void *)ent) + le16_to_cpu(ent->lde_reclen);
419 static inline int lu_dirent_size(struct lu_dirent *ent)
421 if (le16_to_cpu(ent->lde_reclen) == 0) {
422 return (sizeof(*ent) +
423 le16_to_cpu(ent->lde_namelen) + 7) & ~7;
425 return le16_to_cpu(ent->lde_reclen);
428 #define DIR_END_OFF 0xfffffffffffffffeULL
430 struct lustre_handle {
433 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
435 static inline int lustre_handle_is_used(struct lustre_handle *lh)
437 return lh->cookie != 0ull;
440 static inline int lustre_handle_equal(struct lustre_handle *lh1,
441 struct lustre_handle *lh2)
443 return lh1->cookie == lh2->cookie;
446 static inline void lustre_handle_copy(struct lustre_handle *tgt,
447 struct lustre_handle *src)
449 tgt->cookie = src->cookie;
452 #define lustre_msg lustre_msg_v2
453 /* we depend on this structure to be 8-byte aligned */
454 /* this type is only endian-adjusted in lustre_unpack_msg() */
455 struct lustre_msg_v2 {
467 /* without gss, ptlrpc_body is put at the first buffer. */
469 struct lustre_handle pb_handle;
476 __u64 pb_last_committed;
487 extern void lustre_swab_ptlrpc_body(struct ptlrpc_body *pb);
489 /* message body offset for lustre_msg_v2 */
490 /* ptlrpc body offset in all request/reply messages */
491 #define MSG_PTLRPC_BODY_OFF 0
493 /* normal request/reply message record offset */
494 #define REQ_REC_OFF 1
495 #define REPLY_REC_OFF 1
497 /* ldlm request message body offset */
498 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
499 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
501 /* ldlm intent lock message body offset */
502 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
503 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
505 /* ldlm reply message body offset */
506 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
507 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
509 /* Flags that are operation-specific go in the top 16 bits. */
510 #define MSG_OP_FLAG_MASK 0xffff0000
511 #define MSG_OP_FLAG_SHIFT 16
513 /* Flags that apply to all requests are in the bottom 16 bits */
514 #define MSG_GEN_FLAG_MASK 0x0000ffff
515 #define MSG_LAST_REPLAY 1
518 #define MSG_REQ_REPLAY_DONE 8
519 #define MSG_LOCK_REPLAY_DONE 16
522 * Flags for all connect opcodes (MDS_CONNECT, OST_CONNECT)
525 #define MSG_CONNECT_RECOVERING 0x1
526 #define MSG_CONNECT_RECONNECT 0x2
527 #define MSG_CONNECT_REPLAYABLE 0x4
528 //#define MSG_CONNECT_PEER 0x8
529 #define MSG_CONNECT_LIBCLIENT 0x10
530 #define MSG_CONNECT_INITIAL 0x20
531 #define MSG_CONNECT_ASYNC 0x40
532 #define MSG_CONNECT_NEXT_VER 0x80 /* use next version of lustre_msg */
533 #define MSG_CONNECT_TRANSNO 0x100 /* report transno */
536 #define OBD_CONNECT_RDONLY 0x00000001ULL /* client allowed read-only access */
537 #define OBD_CONNECT_INDEX 0x00000002ULL /* connect to specific LOV idx */
538 #define OBD_CONNECT_GRANT 0x00000008ULL /* OSC acquires grant at connect */
539 #define OBD_CONNECT_SRVLOCK 0x00000010ULL /* server takes locks for client */
540 #define OBD_CONNECT_VERSION 0x00000020ULL /* Server supports versions in ocd */
541 #define OBD_CONNECT_REQPORTAL 0x00000040ULL /* Separate portal for non-IO reqs */
542 #define OBD_CONNECT_ACL 0x00000080ULL /* client uses access control lists */
543 #define OBD_CONNECT_XATTR 0x00000100ULL /* client using extended attributes*/
544 #define OBD_CONNECT_TRUNCLOCK 0x00000400ULL /* locks on server for punch b=9528 */
545 #define OBD_CONNECT_IBITS 0x00001000ULL /* support for inodebits locks */
546 #define OBD_CONNECT_JOIN 0x00002000ULL /* files can be concatenated */
547 #define OBD_CONNECT_ATTRFID 0x00004000ULL /* Server supports GetAttr By Fid */
548 #define OBD_CONNECT_NODEVOH 0x00008000ULL /* No open handle for special nodes */
549 #define OBD_CONNECT_LCL_CLIENT 0x00010000ULL /* local 1.8 client */
550 #define OBD_CONNECT_RMT_CLIENT 0x00020000ULL /* Remote 1.8 client */
551 #define OBD_CONNECT_BRW_SIZE 0x00040000ULL /* Max bytes per rpc */
552 #define OBD_CONNECT_QUOTA64 0x00080000ULL /* 64bit qunit_data.qd_count b=10707*/
553 #define OBD_CONNECT_MDS_CAPA 0x00100000ULL /* MDS capability */
554 #define OBD_CONNECT_OSS_CAPA 0x00200000ULL /* OSS capability */
555 #define OBD_CONNECT_CANCELSET 0x00400000ULL /* Early batched cancels. */
556 #define OBD_CONNECT_SOM 0x00800000ULL /* SOM feature */
557 #define OBD_CONNECT_AT 0x01000000ULL /* client uses adaptive timeouts */
558 #define OBD_CONNECT_LRU_RESIZE 0x02000000ULL /* Lru resize feature. */
559 #define OBD_CONNECT_MDS_MDS 0x04000000ULL /* MDS-MDS connection*/
560 #define OBD_CONNECT_REAL 0x08000000ULL /* real connection */
561 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*shrink/enlarge qunit size
563 #define OBD_CONNECT_CKSUM 0x20000000ULL /* support several cksum algos */
564 #define OBD_CONNECT_FID 0x40000000ULL /* FID is supported by server */
566 /* also update obd_connect_names[] for lprocfs_rd_connect_flags()
567 * and lustre/utils/wirecheck.c */
569 #ifdef HAVE_LRU_RESIZE_SUPPORT
570 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
572 #define LRU_RESIZE_CONNECT_FLAG 0
575 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
576 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
577 OBD_CONNECT_IBITS | OBD_CONNECT_JOIN | \
578 OBD_CONNECT_NODEVOH |/* OBD_CONNECT_ATTRFID |*/\
579 OBD_CONNECT_LCL_CLIENT | \
580 OBD_CONNECT_RMT_CLIENT | \
581 OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA | \
582 OBD_CONNECT_MDS_MDS | OBD_CONNECT_CANCELSET | \
584 LRU_RESIZE_CONNECT_FLAG)
585 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
586 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
587 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
588 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_QUOTA64 | \
589 OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET | \
590 OBD_CONNECT_FID | OBD_CONNECT_CKSUM | \
591 LRU_RESIZE_CONNECT_FLAG)
592 #define ECHO_CONNECT_SUPPORTED (0)
593 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_FID)
595 #define MAX_QUOTA_COUNT32 (0xffffffffULL)
597 #define OBD_OCD_VERSION(major,minor,patch,fix) (((major)<<24) + ((minor)<<16) +\
598 ((patch)<<8) + (fix))
599 #define OBD_OCD_VERSION_MAJOR(version) ((int)((version)>>24)&255)
600 #define OBD_OCD_VERSION_MINOR(version) ((int)((version)>>16)&255)
601 #define OBD_OCD_VERSION_PATCH(version) ((int)((version)>>8)&255)
602 #define OBD_OCD_VERSION_FIX(version) ((int)(version)&255)
604 /* This structure is used for both request and reply.
606 * If we eventually have separate connect data for different types, which we
607 * almost certainly will, then perhaps we stick a union in here. */
608 struct obd_connect_data {
609 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
610 __u32 ocd_version; /* lustre release version number */
611 __u32 ocd_grant; /* initial cache grant amount (bytes) */
612 __u32 ocd_index; /* LOV index to connect to */
613 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
614 __u64 ocd_ibits_known; /* inode bits this client understands */
615 __u32 ocd_nllu; /* non-local-lustre-user */
616 __u32 ocd_nllg; /* non-local-lustre-group */
617 __u64 ocd_transno; /* first transno from client to be replayed */
618 __u32 ocd_group; /* MDS group on OST */
619 __u32 ocd_cksum_types; /* supported checksum algorithms */
620 __u64 padding1; /* also fix lustre_swab_connect */
621 __u64 padding2; /* also fix lustre_swab_connect */
624 extern void lustre_swab_connect(struct obd_connect_data *ocd);
627 * Supported checksum algorithms. Up to 32 checksum types are supported.
628 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
629 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
630 * algorithm and also the OBD_FL_CKSUM* flags.
633 OBD_CKSUM_CRC32 = 0x00000001,
634 OBD_CKSUM_ADLER = 0x00000002,
638 * OST requests: OBDO & OBD request records
643 OST_REPLY = 0, /* reply ? */
663 #define OST_FIRST_OPC OST_REPLY
665 typedef uint64_t obd_id;
666 typedef uint64_t obd_gr;
667 typedef uint64_t obd_time;
668 typedef uint64_t obd_size;
669 typedef uint64_t obd_off;
670 typedef uint64_t obd_blocks;
671 typedef uint32_t obd_blksize;
672 typedef uint32_t obd_mode;
673 typedef uint32_t obd_uid;
674 typedef uint32_t obd_gid;
675 typedef uint32_t obd_flag;
676 typedef uint64_t obd_valid;
677 typedef uint32_t obd_count;
679 #define OBD_FL_INLINEDATA (0x00000001)
680 #define OBD_FL_OBDMDEXISTS (0x00000002)
681 #define OBD_FL_DELORPHAN (0x00000004) /* if set in o_flags delete orphans */
682 #define OBD_FL_NORPC (0x00000008) /* set in o_flags do in OSC not OST */
683 #define OBD_FL_IDONLY (0x00000010) /* set in o_flags only adjust obj id*/
684 #define OBD_FL_RECREATE_OBJS (0x00000020) /* recreate missing obj */
685 #define OBD_FL_DEBUG_CHECK (0x00000040) /* echo client/server debug check */
686 #define OBD_FL_NO_USRQUOTA (0x00000100) /* the object's owner is over quota */
687 #define OBD_FL_NO_GRPQUOTA (0x00000200) /* the object's group is over quota */
690 * Set this to delegate DLM locking during obd_punch() to the OSTs. Only OSTs
691 * that declared OBD_CONNECT_TRUNCLOCK in their connect flags support this
694 #define OBD_FL_TRUNCLOCK (0x00000800)
699 #define OBD_FL_CKSUM_CRC32 (0x00001000)
700 #define OBD_FL_CKSUM_ADLER (0x00002000)
701 #define OBD_FL_CKSUM_ALL (OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER)
704 * This should not be smaller than sizeof(struct lustre_handle) + sizeof(struct
705 * llog_cookie) + sizeof(struct ll_fid). Nevertheless struct ll_fid is not
706 * longer stored in o_inline, we keep this just for case.
708 #define OBD_INLINESZ 80
710 /* Note: 64-bit types are 64-bit aligned in structure */
712 obd_valid o_valid; /* hot fields in this obdo */
716 obd_size o_size; /* o_size-o_blocks == ost_lvb */
720 obd_blocks o_blocks; /* brw: cli sent cached bytes */
723 /* 32-bit fields start here: keep an even number of them via padding */
724 obd_blksize o_blksize; /* optimal IO blocksize */
725 obd_mode o_mode; /* brw: cli sent cache remain */
729 obd_count o_nlink; /* brw: checksum */
730 obd_count o_generation;
731 obd_count o_misc; /* brw: o_dropped */
732 __u32 o_easize; /* epoch in ost writes */
734 __u32 o_stripe_idx; /* holds stripe idx */
736 char o_inline[OBD_INLINESZ];
737 /* lustre_handle + llog_cookie */
740 #define o_dirty o_blocks
741 #define o_undirty o_mode
742 #define o_dropped o_misc
743 #define o_cksum o_nlink
745 extern void lustre_swab_obdo (struct obdo *o);
748 #define LOV_MAGIC_V1 0x0BD10BD0
749 #define LOV_MAGIC LOV_MAGIC_V1
750 #define LOV_MAGIC_JOIN 0x0BD20BD0
752 #define LOV_PATTERN_RAID0 0x001 /* stripes are used round-robin */
753 #define LOV_PATTERN_RAID1 0x002 /* stripes are mirrors of each other */
754 #define LOV_PATTERN_FIRST 0x100 /* first stripe is not in round-robin */
755 #define LOV_PATTERN_CMOBD 0x200
757 #define LOV_OBJECT_GROUP_DEFAULT ~0ULL
758 #define LOV_OBJECT_GROUP_CLEAR 0ULL
760 #define lov_ost_data lov_ost_data_v1
761 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
762 __u64 l_object_id; /* OST object ID */
763 __u64 l_object_gr; /* OST object group (creating MDS number) */
764 __u32 l_ost_gen; /* generation of this l_ost_idx */
765 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
768 #define lov_mds_md lov_mds_md_v1
769 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
770 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
771 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
772 __u64 lmm_object_id; /* LOV object ID */
773 __u64 lmm_object_gr; /* LOV object group */
774 __u32 lmm_stripe_size; /* size of stripe in bytes */
775 __u32 lmm_stripe_count; /* num stripes in use for this object */
776 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
779 extern void lustre_swab_lov_mds_md(struct lov_mds_md *llm);
781 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
782 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
784 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
785 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
786 #define XATTR_NAME_LOV "trusted.lov"
788 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
789 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
790 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
791 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
792 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
793 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
794 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
795 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
796 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
797 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
798 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
799 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
800 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
801 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
802 #define OBD_MD_FLINLINE (0x00008000ULL) /* inline data */
803 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
804 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
805 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
806 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file handle */
807 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
808 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
809 #define OBD_MD_FLOSCOPQ (0x00400000ULL) /* osc opaque data */
810 #define OBD_MD_FLCOOKIE (0x00800000ULL) /* log cancellation cookie */
811 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
812 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
813 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write easize is epoch */
814 /* ->mds if epoch opens or closes */
815 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
816 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
817 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
818 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
819 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
821 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
822 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
823 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
825 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
826 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
827 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
828 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
829 #define OBD_MD_FLRMTPERM (0x0000010000000000ULL) /* remote permission */
830 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
831 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
832 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
833 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
835 #define OBD_MD_FLRMTLSETFACL (0x0001000000000000ULL) /* lfs lsetfacl case */
836 #define OBD_MD_FLRMTLGETFACL (0x0002000000000000ULL) /* lfs lgetfacl case */
837 #define OBD_MD_FLRMTRSETFACL (0x0004000000000000ULL) /* lfs rsetfacl case */
838 #define OBD_MD_FLRMTRGETFACL (0x0008000000000000ULL) /* lfs rgetfacl case */
840 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
841 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
842 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
843 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
844 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP)
846 static inline struct lustre_handle *obdo_handle(struct obdo *oa)
848 return (struct lustre_handle *)oa->o_inline;
851 static inline struct llog_cookie *obdo_logcookie(struct obdo *oa)
853 return (struct llog_cookie *)(oa->o_inline +
854 sizeof(struct lustre_handle));
856 /* don't forget obdo_fid which is way down at the bottom so it can
857 * come after the definition of llog_cookie */
870 __u32 os_state; /* positive error code on server */
882 extern void lustre_swab_obd_statfs (struct obd_statfs *os);
883 #define OBD_STATFS_NODELAY 0x0001 /* requests should be send without delay
884 * and resends for avoid deadlocks */
886 /* ost_body.data values for OST_BRW */
888 #define OBD_BRW_READ 0x01
889 #define OBD_BRW_WRITE 0x02
890 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
891 #define OBD_BRW_SYNC 0x08
892 #define OBD_BRW_CHECK 0x10
893 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
894 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
895 #define OBD_BRW_DROP 0x80 /* drop the page after IO */
896 #define OBD_BRW_NOQUOTA 0x100
897 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
899 #define OBD_OBJECT_EOF 0xffffffffffffffffULL
901 #define OST_MIN_PRECREATE 32
902 #define OST_MAX_PRECREATE 20000
911 extern void lustre_swab_obd_ioobj (struct obd_ioobj *ioo);
913 /* multiple of 8 bytes => can array */
914 struct niobuf_remote {
920 extern void lustre_swab_niobuf_remote (struct niobuf_remote *nbr);
922 /* request structure for OST's */
928 extern void lustre_swab_ost_body (struct ost_body *b);
929 extern void lustre_swab_ost_last_id(obd_id *id);
931 /* lock value block communicated between the filter and llite */
933 /* OST_LVB_ERR_INIT is needed because the return code in rc is
934 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
935 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
936 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
937 #define OST_LVB_IS_ERR(blocks) \
938 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
939 #define OST_LVB_SET_ERR(blocks, rc) \
940 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
941 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
951 extern void lustre_swab_ost_lvb(struct ost_lvb *);
960 MDS_GETATTR_NAME = 34,
971 MDS_DONE_WRITING = 45,
976 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
982 #define MDS_FIRST_OPC MDS_GETATTR
999 } mds_reint_t, mdt_reint_t;
1001 extern void lustre_swab_generic_32s (__u32 *val);
1003 /* the disposition of the intent outlines what was executed */
1004 #define DISP_IT_EXECD 0x00000001
1005 #define DISP_LOOKUP_EXECD 0x00000002
1006 #define DISP_LOOKUP_NEG 0x00000004
1007 #define DISP_LOOKUP_POS 0x00000008
1008 #define DISP_OPEN_CREATE 0x00000010
1009 #define DISP_OPEN_OPEN 0x00000020
1010 #define DISP_ENQ_COMPLETE 0x00400000
1011 #define DISP_ENQ_OPEN_REF 0x00800000
1012 #define DISP_ENQ_CREATE_REF 0x01000000
1013 #define DISP_OPEN_LOCK 0x02000000
1015 /* INODE LOCK PARTS */
1016 #define MDS_INODELOCK_LOOKUP 0x000001 /* dentry, mode, owner, group */
1017 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1018 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1020 /* Do not forget to increase MDS_INODELOCK_MAXSHIFT when adding new bits */
1021 #define MDS_INODELOCK_MAXSHIFT 2
1022 /* This FULL lock is useful to take on unlink sort of operations */
1023 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1026 __u64 id; /* holds object id */
1027 __u32 generation; /* holds object generation */
1029 __u32 f_type; /* holds object type or stripe idx when passing it to
1030 * OST for saving into EA. */
1033 extern void lustre_swab_ll_fid (struct ll_fid *fid);
1035 #define MDS_STATUS_CONN 1
1036 #define MDS_STATUS_LOV 2
1038 struct mds_status_req {
1043 extern void lustre_swab_mds_status_req (struct mds_status_req *r);
1045 /* mdt_thread_info.mti_flags. */
1047 /* The flag indicates Size-on-MDS attributes are changed. */
1048 MF_SOM_CHANGE = (1 << 0),
1049 /* Flags indicates an epoch opens or closes. */
1050 MF_EPOCH_OPEN = (1 << 1),
1051 MF_EPOCH_CLOSE = (1 << 2),
1052 MF_MDC_CANCEL_FID1 = (1 << 3),
1053 MF_MDC_CANCEL_FID2 = (1 << 4),
1054 MF_MDC_CANCEL_FID3 = (1 << 5),
1055 MF_MDC_CANCEL_FID4 = (1 << 6),
1058 #define MF_SOM_LOCAL_FLAGS (MF_MDC_CANCEL_FID1 | MF_MDC_CANCEL_FID2 | \
1059 MF_MDC_CANCEL_FID3 | MF_MDC_CANCEL_FID4)
1061 #define MDS_BFLAG_UNCOMMITTED_WRITES 0x1
1062 #define MDS_BFLAG_EXT_FLAGS 0x80000000 /* == EXT3_RESERVED_FL */
1064 /* these should be identical to their EXT3_*_FL counterparts, and are
1065 * redefined here only to avoid dragging in ext3_fs.h */
1066 #define MDS_SYNC_FL 0x00000008 /* Synchronous updates */
1067 #define MDS_IMMUTABLE_FL 0x00000010 /* Immutable file */
1068 #define MDS_APPEND_FL 0x00000020 /* writes to file may only append */
1069 #define MDS_NOATIME_FL 0x00000080 /* do not update atime */
1070 #define MDS_DIRSYNC_FL 0x00010000 /* dirsync behaviour (dir only) */
1073 /* If MDS_BFLAG_IOC_FLAGS is set it means we requested EXT3_*_FL inode flags
1074 * and we need to decode these into local S_* flags in the inode. Otherwise
1075 * we pass flags straight through (see bug 9486). */
1076 static inline int ll_ext_to_inode_flags(int flags)
1078 return (flags & MDS_BFLAG_EXT_FLAGS) ?
1079 (((flags & MDS_SYNC_FL) ? S_SYNC : 0) |
1080 ((flags & MDS_NOATIME_FL) ? S_NOATIME : 0) |
1081 ((flags & MDS_APPEND_FL) ? S_APPEND : 0) |
1082 #if defined(S_DIRSYNC)
1083 ((flags & MDS_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1085 ((flags & MDS_IMMUTABLE_FL) ? S_IMMUTABLE : 0)) :
1086 (flags & ~MDS_BFLAG_EXT_FLAGS);
1089 /* If MDS_BFLAG_EXT_FLAGS is set it means we requested EXT3_*_FL inode flags
1090 * and we pass these straight through. Otherwise we need to convert from
1091 * S_* flags to their EXT3_*_FL equivalents (see bug 9486). */
1092 static inline int ll_inode_to_ext_flags(int oflags, int iflags)
1094 return (oflags & MDS_BFLAG_EXT_FLAGS) ? (oflags & ~MDS_BFLAG_EXT_FLAGS):
1095 (((iflags & S_SYNC) ? MDS_SYNC_FL : 0) |
1096 ((iflags & S_NOATIME) ? MDS_NOATIME_FL : 0) |
1097 ((iflags & S_APPEND) ? MDS_APPEND_FL : 0) |
1098 #if defined(S_DIRSYNC)
1099 ((iflags & S_DIRSYNC) ? MDS_DIRSYNC_FL : 0) |
1101 ((iflags & S_IMMUTABLE) ? MDS_IMMUTABLE_FL : 0));
1108 struct lustre_handle handle;
1110 __u64 size; /* Offset, in the case of MDS_READPAGE */
1114 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1116 __u64 ino; /* for 1.6 compatibility */
1123 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1125 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1126 __u32 generation; /* for 1.6 compatibility */
1131 __u32 max_cookiesize;
1132 __u32 padding_4; /* also fix lustre_swab_mdt_body */
1138 struct lustre_handle handle;
1140 __u64 size; /* Offset, in the case of MDS_READPAGE */
1144 __u64 blocks; /* XID, in the case of MDS_READPAGE */
1153 __u32 flags; /* from vfs for pin/unpin, MDS_BFLAG for close */
1155 __u32 nlink; /* #bytes to read in the case of MDS_READPAGE */
1161 __u32 max_cookiesize;
1162 __u32 padding_4; /* also fix lustre_swab_mds_body */
1165 extern void lustre_swab_mds_body (struct mds_body *b);
1166 extern void lustre_swab_mdt_body (struct mdt_body *b);
1169 struct lustre_handle handle;
1175 extern void lustre_swab_mdt_epoch (struct mdt_epoch *b);
1177 #define Q_QUOTACHECK 0x800100
1178 #define Q_INITQUOTA 0x800101 /* init slave limits */
1179 #define Q_GETOINFO 0x800102 /* get obd quota info */
1180 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1182 #define Q_TYPESET(oqc, type) \
1183 ((oqc)->qc_type == type || (oqc)->qc_type == UGQUOTA)
1185 #define Q_GETOCMD(oqc) \
1186 ((oqc)->qc_cmd == Q_GETOINFO || (oqc)->qc_cmd == Q_GETOQUOTA)
1188 struct obd_quotactl {
1193 struct obd_dqinfo qc_dqinfo;
1194 struct obd_dqblk qc_dqblk;
1197 extern void lustre_swab_obd_quotactl(struct obd_quotactl *q);
1199 /* inode access permission for remote user, the inode info are omitted,
1200 * for client knows them. */
1201 struct mds_remote_perm {
1206 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1209 /* permissions for md_perm.mp_perm */
1211 CFS_SETUID_PERM = 0x01,
1212 CFS_SETGID_PERM = 0x02,
1213 CFS_SETGRP_PERM = 0x04,
1214 CFS_RMTACL_PERM = 0x08
1217 extern void lustre_swab_mds_remote_perm(struct mds_remote_perm *p);
1219 struct mdt_remote_perm {
1224 __u32 rp_access_perm; /* MAY_READ/WRITE/EXEC */
1227 extern void lustre_swab_mdt_remote_perm(struct mdt_remote_perm *p);
1229 struct mds_rec_setattr {
1236 struct ll_fid sa_fid;
1237 __u64 sa_valid; /* MDS_ATTR_* attributes */
1244 __u32 sa_attr_flags;
1245 __u32 sa_padding; /* also fix lustre_swab_mds_rec_setattr */
1248 extern void lustre_swab_mds_rec_setattr (struct mds_rec_setattr *sa);
1250 struct mdt_rec_setattr {
1257 struct lu_fid sa_fid;
1266 __u32 sa_attr_flags;
1274 extern void lustre_swab_mdt_rec_setattr (struct mdt_rec_setattr *sa);
1277 * Attribute flags used in mds_rec_setattr::sa_valid.
1278 * The kernel's #defines for ATTR_* should not be used over the network
1279 * since the client and MDS may run different kernels (see bug 13828)
1280 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1282 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1283 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1284 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1285 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1286 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1287 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1288 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1289 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1290 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1291 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1292 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1293 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1294 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1295 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1296 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1297 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1298 #define MDS_ATTR_TRUNC 0x10000ULL /* = 65536 */
1300 #define ATTR_TRUNC MDS_ATTR_TRUNC
1303 #define FMODE_READ 00000001
1304 #define FMODE_WRITE 00000002
1307 #define FMODE_EPOCH 01000000
1308 #define FMODE_EPOCHLCK 02000000
1309 #define FMODE_SOM 04000000
1310 #define FMODE_CLOSED 0
1312 #define MDS_OPEN_CREATED 00000010
1313 #define MDS_OPEN_CROSS 00000020
1315 #define MDS_FMODE_EXEC 00000004
1316 #define MDS_OPEN_CREAT 00000100
1317 #define MDS_OPEN_EXCL 00000200
1318 #define MDS_OPEN_TRUNC 00001000
1319 #define MDS_OPEN_APPEND 00002000
1320 #define MDS_OPEN_SYNC 00010000
1321 #define MDS_OPEN_DIRECTORY 00200000
1323 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1324 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1325 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file*/
1326 #define MDS_CREATE_RMT_ACL 01000000000 /* indicate create on remote server
1327 * with default ACL */
1328 #define MDS_CREATE_SLAVE_OBJ 02000000000 /* indicate create slave object
1329 * actually, this is for create, not
1330 * conflict with other open flags */
1331 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1332 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1333 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1335 /* permission for create non-directory file */
1336 #define MAY_CREATE (1 << 7)
1337 /* permission for create directory file */
1338 #define MAY_LINK (1 << 8)
1339 /* permission for delete from the directory */
1340 #define MAY_UNLINK (1 << 9)
1341 /* source's permission for rename */
1342 #define MAY_RENAME_SRC (1 << 10)
1343 /* target's permission for rename */
1344 #define MAY_RENAME_TAR (1 << 11)
1345 /* part (parent's) VTX permission check */
1346 #define MAY_VTX_PART (1 << 12)
1347 /* full VTX permission check */
1348 #define MAY_VTX_FULL (1 << 13)
1349 /* lfs rgetfacl permission check */
1350 #define MAY_RGETFACL (1 << 14)
1353 MDS_CHECK_SPLIT = 1 << 0,
1354 MDS_CROSS_REF = 1 << 1,
1355 MDS_VTX_BYPASS = 1 << 2,
1356 MDS_PERM_BYPASS = 1 << 3
1359 struct mds_rec_join {
1360 struct ll_fid jr_fid;
1364 extern void lustre_swab_mds_rec_join (struct mds_rec_join *jr);
1366 struct mdt_rec_join {
1367 struct lu_fid jr_fid;
1371 extern void lustre_swab_mdt_rec_join (struct mdt_rec_join *jr);
1373 struct mds_rec_create {
1378 __u32 cr_flags; /* for use with open */
1380 struct ll_fid cr_fid;
1381 struct ll_fid cr_replayfid;
1385 __u32 cr_padding_1; /* also fix lustre_swab_mds_rec_create */
1386 __u32 cr_padding_2; /* also fix lustre_swab_mds_rec_create */
1387 __u32 cr_padding_3; /* also fix lustre_swab_mds_rec_create */
1388 __u32 cr_padding_4; /* also fix lustre_swab_mds_rec_create */
1389 __u32 cr_padding_5; /* also fix lustre_swab_mds_rec_create */
1392 extern void lustre_swab_mds_rec_create (struct mds_rec_create *cr);
1394 struct mdt_rec_create {
1401 struct lu_fid cr_fid1;
1402 struct lu_fid cr_fid2;
1403 struct lustre_handle cr_old_handle; /* u64 handle in case of open replay */
1407 __u64 cr_padding_1; /* pad for 64 bits*/
1410 __u32 cr_flags; /* for use with open */
1416 extern void lustre_swab_mdt_rec_create (struct mdt_rec_create *cr);
1418 struct mds_rec_link {
1425 struct ll_fid lk_fid1;
1426 struct ll_fid lk_fid2;
1428 __u32 lk_padding_1; /* also fix lustre_swab_mds_rec_link */
1429 __u32 lk_padding_2; /* also fix lustre_swab_mds_rec_link */
1430 __u32 lk_padding_3; /* also fix lustre_swab_mds_rec_link */
1431 __u32 lk_padding_4; /* also fix lustre_swab_mds_rec_link */
1434 extern void lustre_swab_mds_rec_link (struct mds_rec_link *lk);
1436 struct mdt_rec_link {
1443 struct lu_fid lk_fid1;
1444 struct lu_fid lk_fid2;
1458 struct mds_rec_unlink {
1465 struct ll_fid ul_fid1;
1466 struct ll_fid ul_fid2;
1468 __u32 ul_padding_1; /* also fix lustre_swab_mds_rec_unlink */
1469 __u32 ul_padding_2; /* also fix lustre_swab_mds_rec_unlink */
1470 __u32 ul_padding_3; /* also fix lustre_swab_mds_rec_unlink */
1471 __u32 ul_padding_4; /* also fix lustre_swab_mds_rec_unlink */
1474 extern void lustre_swab_mds_rec_unlink (struct mds_rec_unlink *ul);
1476 struct mdt_rec_unlink {
1483 struct lu_fid ul_fid1;
1484 struct lu_fid ul_fid2;
1498 struct mds_rec_rename {
1505 struct ll_fid rn_fid1;
1506 struct ll_fid rn_fid2;
1508 __u32 rn_padding_1; /* also fix lustre_swab_mds_rec_rename */
1509 __u32 rn_padding_2; /* also fix lustre_swab_mds_rec_rename */
1510 __u32 rn_padding_3; /* also fix lustre_swab_mds_rec_rename */
1511 __u32 rn_padding_4; /* also fix lustre_swab_mds_rec_rename */
1514 extern void lustre_swab_mds_rec_rename (struct mds_rec_rename *rn);
1516 struct mdt_rec_rename {
1523 struct lu_fid rn_fid1;
1524 struct lu_fid rn_fid2;
1530 __u32 rn_bias; /* some operation flags */
1531 __u32 rn_mode; /* cross-ref rename has mode */
1538 struct mdt_rec_setxattr {
1545 struct lu_fid sx_fid;
1546 __u64 sx_padding_1; /* These three members are lu_fid size */
1558 __u32 sx_padding_10;
1559 __u32 sx_padding_11;
1562 struct mdt_rec_reint {
1569 struct lu_fid rr_fid1;
1570 struct lu_fid rr_fid2;
1578 __u32 rr_padding_1; /* also fix lustre_swab_mdt_rec_reint */
1579 __u32 rr_padding_2; /* also fix lustre_swab_mdt_rec_reint */
1580 __u32 rr_padding_3; /* also fix lustre_swab_mdt_rec_reint */
1581 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
1584 extern void lustre_swab_mdt_rec_reint(struct mdt_rec_reint *rr);
1587 __u32 ld_tgt_count; /* how many MDS's */
1588 __u32 ld_active_tgt_count; /* how many active */
1589 struct obd_uuid ld_uuid;
1592 extern void lustre_swab_lmv_desc (struct lmv_desc *ld);
1599 extern void lustre_swab_md_fld (struct md_fld *mf);
1604 FLD_FIRST_OPC = FLD_QUERY
1610 SEQ_FIRST_OPC = SEQ_QUERY
1614 SEQ_ALLOC_SUPER = 0,
1619 * LOV data structures
1622 #define LOV_MIN_STRIPE_SIZE 65536 /* maximum PAGE_SIZE (ia64), power of 2 */
1623 #define LOV_MAX_STRIPE_COUNT 160 /* until bug 4424 is fixed */
1624 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
1626 #define LOV_MAX_UUID_BUFFER_SIZE 8192
1627 /* The size of the buffer the lov/mdc reserves for the
1628 * array of UUIDs returned by the MDS. With the current
1629 * protocol, this will limit the max number of OSTs per LOV */
1631 #define LOV_DESC_MAGIC 0xB0CCDE5C
1633 /* LOV settings descriptor (should only contain static info) */
1635 __u32 ld_tgt_count; /* how many OBD's */
1636 __u32 ld_active_tgt_count; /* how many active */
1637 __u32 ld_default_stripe_count; /* how many objects are used */
1638 __u32 ld_pattern; /* default PATTERN_RAID0 */
1639 __u64 ld_default_stripe_size; /* in bytes */
1640 __u64 ld_default_stripe_offset; /* in bytes */
1641 __u32 ld_padding_0; /* unused */
1642 __u32 ld_qos_maxage; /* in second */
1643 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
1644 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
1645 struct obd_uuid ld_uuid;
1648 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
1650 extern void lustre_swab_lov_desc (struct lov_desc *ld);
1655 /* opcodes -- MUST be distinct from OST/MDS opcodes */
1660 LDLM_BL_CALLBACK = 104,
1661 LDLM_CP_CALLBACK = 105,
1662 LDLM_GL_CALLBACK = 106,
1665 #define LDLM_FIRST_OPC LDLM_ENQUEUE
1667 #define RES_NAME_SIZE 4
1668 struct ldlm_res_id {
1669 __u64 name[RES_NAME_SIZE];
1672 extern void lustre_swab_ldlm_res_id (struct ldlm_res_id *id);
1687 #define LCK_MODE_NUM 7
1697 #define LDLM_MIN_TYPE LDLM_PLAIN
1699 struct ldlm_extent {
1705 static inline int ldlm_extent_overlap(struct ldlm_extent *ex1,
1706 struct ldlm_extent *ex2)
1708 return (ex1->start <= ex2->end) && (ex2->start <= ex1->end);
1711 struct ldlm_inodebits {
1718 __u64 blocking_export; /* not actually used over the wire */
1719 __u32 blocking_pid; /* not actually used over the wire */
1723 /* it's important that the fields of the ldlm_extent structure match
1724 * the first fields of the ldlm_flock structure because there is only
1725 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
1726 * this ever changes we will need to swab the union differently based
1727 * on the resource type. */
1730 struct ldlm_extent l_extent;
1731 struct ldlm_flock l_flock;
1732 struct ldlm_inodebits l_inodebits;
1733 } ldlm_policy_data_t;
1735 extern void lustre_swab_ldlm_policy_data (ldlm_policy_data_t *d);
1737 struct ldlm_intent {
1741 extern void lustre_swab_ldlm_intent (struct ldlm_intent *i);
1743 struct ldlm_resource_desc {
1744 ldlm_type_t lr_type;
1745 __u32 lr_padding; /* also fix lustre_swab_ldlm_resource_desc */
1746 struct ldlm_res_id lr_name;
1749 extern void lustre_swab_ldlm_resource_desc (struct ldlm_resource_desc *r);
1751 struct ldlm_lock_desc {
1752 struct ldlm_resource_desc l_resource;
1753 ldlm_mode_t l_req_mode;
1754 ldlm_mode_t l_granted_mode;
1755 ldlm_policy_data_t l_policy_data;
1758 extern void lustre_swab_ldlm_lock_desc (struct ldlm_lock_desc *l);
1760 #define LDLM_LOCKREQ_HANDLES 2
1761 #define LDLM_ENQUEUE_CANCEL_OFF 1
1763 struct ldlm_request {
1766 struct ldlm_lock_desc lock_desc;
1767 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
1770 extern void lustre_swab_ldlm_request (struct ldlm_request *rq);
1772 /* If LDLM_ENQUEUE, 1 slot is already occupied, 1 is available.
1773 * Otherwise, 2 are available. */
1774 #define ldlm_request_bufsize(count,type) \
1776 int _avail = LDLM_LOCKREQ_HANDLES; \
1777 _avail -= (type == LDLM_ENQUEUE ? LDLM_ENQUEUE_CANCEL_OFF : 0); \
1778 sizeof(struct ldlm_request) + \
1779 (count > _avail ? count - _avail : 0) * \
1780 sizeof(struct lustre_handle); \
1785 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
1786 struct ldlm_lock_desc lock_desc;
1787 struct lustre_handle lock_handle;
1788 __u64 lock_policy_res1;
1789 __u64 lock_policy_res2;
1792 extern void lustre_swab_ldlm_reply (struct ldlm_reply *r);
1795 * Opcodes for mountconf (mgs and mgc)
1800 MGS_EXCEPTION, /* node died, etc. */
1801 MGS_TARGET_REG, /* whenever target starts up */
1806 #define MGS_FIRST_OPC MGS_CONNECT
1808 #define MGS_PARAM_MAXLEN 1024
1809 #define KEY_SET_INFO "set_info"
1811 struct mgs_send_param {
1812 char mgs_param[MGS_PARAM_MAXLEN];
1815 /* We pass this info to the MGS so it can write config logs */
1816 #define MTI_NAME_MAXLEN 64
1817 #define MTI_PARAM_MAXLEN 4096
1818 #define MTI_NIDS_MAX 32
1819 struct mgs_target_info {
1820 __u32 mti_lustre_ver;
1821 __u32 mti_stripe_index;
1822 __u32 mti_config_ver;
1824 __u32 mti_nid_count;
1825 __u32 padding; /* 64 bit align */
1826 char mti_fsname[MTI_NAME_MAXLEN];
1827 char mti_svname[MTI_NAME_MAXLEN];
1828 char mti_uuid[sizeof(struct obd_uuid)];
1829 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t)*/
1830 char mti_params[MTI_PARAM_MAXLEN];
1833 extern void lustre_swab_mgs_target_info(struct mgs_target_info *oinfo);
1835 /* Config marker flags (in config log) */
1836 #define CM_START 0x01
1838 #define CM_SKIP 0x04
1839 #define CM_UPGRADE146 0x08
1840 #define CM_EXCLUDE 0x10
1841 #define CM_START_SKIP (CM_START | CM_SKIP)
1844 __u32 cm_step; /* aka config version */
1846 __u32 cm_vers; /* lustre release version number */
1847 __u32 padding; /* 64 bit align */
1848 time_t cm_createtime; /*when this record was first created */
1849 time_t cm_canceltime; /*when this record is no longer valid*/
1850 char cm_tgtname[MTI_NAME_MAXLEN];
1851 char cm_comment[MTI_NAME_MAXLEN];
1855 * Opcodes for multiple servers.
1864 #define OBD_FIRST_OPC OBD_PING
1866 /* catalog of log objects */
1868 /* Identifier for a single log object */
1873 } __attribute__((packed));
1875 /* Records written to the CATALOGS list */
1876 #define CATLIST "CATALOGS"
1878 struct llog_logid lci_logid;
1882 } __attribute__((packed));
1884 /*join file lov mds md*/
1885 struct lov_mds_md_join {
1886 struct lov_mds_md lmmj_md;
1887 /*join private info*/
1888 struct llog_logid lmmj_array_id; /*array object id*/
1889 __u32 lmmj_extent_count; /*array extent count*/
1892 /* Log data record types - there is no specific reason that these need to
1893 * be related to the RPC opcodes, but no reason not to (may be handy later?)
1895 #define LLOG_OP_MAGIC 0x10600000
1896 #define LLOG_OP_MASK 0xfff00000
1899 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
1900 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
1901 OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000,
1902 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_UNLINK,
1903 MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) | REINT_SETATTR,
1904 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
1905 PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, /* obsolete */
1906 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
1907 LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000,
1908 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
1909 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
1913 * for now, continue to support old pad records which have 0 for their
1914 * type but still need to be swabbed for their length
1916 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
1917 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == \
1918 __swab32(LLOG_OP_MAGIC) || \
1919 (((r)->lrh_type == 0) && ((r)->lrh_len > LLOG_CHUNK_SIZE)))
1921 /* Log record header - stored in little endian order.
1922 * Each record must start with this struct, end with a llog_rec_tail,
1923 * and be a multiple of 256 bits in size.
1925 struct llog_rec_hdr {
1932 struct llog_rec_tail {
1937 struct llog_logid_rec {
1938 struct llog_rec_hdr lid_hdr;
1939 struct llog_logid lid_id;
1945 struct llog_rec_tail lid_tail;
1946 } __attribute__((packed));
1948 /* MDS extent description
1949 * It is for joined file extent info, each extent info for joined file
1950 * just like (start, end, lmm).
1952 struct mds_extent_desc {
1953 __u64 med_start; /* extent start */
1954 __u64 med_len; /* extent length */
1955 struct lov_mds_md med_lmm; /* extent's lmm */
1957 /*Joined file array extent log record*/
1958 struct llog_array_rec {
1959 struct llog_rec_hdr lmr_hdr;
1960 struct mds_extent_desc lmr_med;
1961 struct llog_rec_tail lmr_tail;
1964 struct llog_create_rec {
1965 struct llog_rec_hdr lcr_hdr;
1966 struct ll_fid lcr_fid;
1970 struct llog_rec_tail lcr_tail;
1971 } __attribute__((packed));
1973 struct llog_orphan_rec {
1974 struct llog_rec_hdr lor_hdr;
1978 struct llog_rec_tail lor_tail;
1979 } __attribute__((packed));
1981 struct llog_unlink_rec {
1982 struct llog_rec_hdr lur_hdr;
1986 struct llog_rec_tail lur_tail;
1987 } __attribute__((packed));
1989 struct llog_setattr_rec {
1990 struct llog_rec_hdr lsr_hdr;
1996 struct llog_rec_tail lsr_tail;
1997 } __attribute__((packed));
1999 struct llog_size_change_rec {
2000 struct llog_rec_hdr lsc_hdr;
2001 struct ll_fid lsc_fid;
2004 struct llog_rec_tail lsc_tail;
2005 } __attribute__((packed));
2010 } __attribute__((packed));
2012 struct llog_gen_rec {
2013 struct llog_rec_hdr lgr_hdr;
2014 struct llog_gen lgr_gen;
2015 struct llog_rec_tail lgr_tail;
2017 /* On-disk header structure of each log object, stored in little endian order */
2018 #define LLOG_CHUNK_SIZE 8192
2019 #define LLOG_HEADER_SIZE (96)
2020 #define LLOG_BITMAP_BYTES (LLOG_CHUNK_SIZE - LLOG_HEADER_SIZE)
2022 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2024 /* flags for the logs */
2025 #define LLOG_F_ZAP_WHEN_EMPTY 0x1
2026 #define LLOG_F_IS_CAT 0x2
2027 #define LLOG_F_IS_PLAIN 0x4
2029 struct llog_log_hdr {
2030 struct llog_rec_hdr llh_hdr;
2031 __u64 llh_timestamp;
2033 __u32 llh_bitmap_offset;
2037 /* for a catalog the first plain slot is next to it */
2038 struct obd_uuid llh_tgtuuid;
2039 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32) - 23];
2040 __u32 llh_bitmap[LLOG_BITMAP_BYTES/sizeof(__u32)];
2041 struct llog_rec_tail llh_tail;
2042 } __attribute__((packed));
2044 #define LLOG_BITMAP_SIZE(llh) ((llh->llh_hdr.lrh_len - \
2045 llh->llh_bitmap_offset - \
2046 sizeof(llh->llh_tail)) * 8)
2048 /* log cookies are used to reference a specific log file and a record therein */
2049 struct llog_cookie {
2050 struct llog_logid lgc_lgl;
2054 } __attribute__((packed));
2057 enum llogd_rpc_ops {
2058 LLOG_ORIGIN_HANDLE_CREATE = 501,
2059 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2060 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2061 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2062 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2063 LLOG_ORIGIN_CONNECT = 506,
2064 LLOG_CATINFO = 507, /* for lfs catinfo */
2065 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2066 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2068 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2072 struct llog_logid lgd_logid;
2074 __u32 lgd_llh_flags;
2076 __u32 lgd_saved_index;
2078 __u64 lgd_cur_offset;
2079 } __attribute__((packed));
2081 struct llogd_conn_body {
2082 struct llog_gen lgdc_gen;
2083 struct llog_logid lgdc_logid;
2084 __u32 lgdc_ctxt_idx;
2085 } __attribute__((packed));
2087 struct lov_user_ost_data_join { /* per-stripe data structure */
2088 __u64 l_extent_start; /* extent start*/
2089 __u64 l_extent_end; /* extent end*/
2090 __u64 l_object_id; /* OST object ID */
2091 __u64 l_object_gr; /* OST object group (creating MDS number) */
2092 __u32 l_ost_gen; /* generation of this OST index */
2093 __u32 l_ost_idx; /* OST index in LOV */
2094 } __attribute__((packed));
2096 struct lov_user_md_join { /* LOV EA user data (host-endian) */
2097 __u32 lmm_magic; /* magic number = LOV_MAGIC_JOIN */
2098 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
2099 __u64 lmm_object_id; /* LOV object ID */
2100 __u64 lmm_object_gr; /* LOV object group */
2101 __u32 lmm_stripe_size; /* size of stripe in bytes */
2102 __u32 lmm_stripe_count; /* num stripes in use for this object */
2103 __u32 lmm_extent_count; /* extent count of lmm*/
2104 __u64 lmm_tree_id; /* mds tree object id */
2105 __u64 lmm_tree_gen; /* mds tree object gen */
2106 struct llog_logid lmm_array_id; /* mds extent desc llog object id */
2107 struct lov_user_ost_data_join lmm_objects[0]; /* per-stripe data */
2108 } __attribute__((packed));
2110 extern void lustre_swab_lov_user_md(struct lov_user_md *lum);
2111 extern void lustre_swab_lov_user_md_objects(struct lov_user_md *lum);
2112 extern void lustre_swab_lov_user_md_join(struct lov_user_md_join *lumj);
2115 extern void lustre_swab_llogd_body (struct llogd_body *d);
2116 extern void lustre_swab_llog_hdr (struct llog_log_hdr *h);
2117 extern void lustre_swab_llogd_conn_body (struct llogd_conn_body *d);
2118 extern void lustre_swab_llog_rec(struct llog_rec_hdr *rec,
2119 struct llog_rec_tail *tail);
2122 extern void lustre_swab_lustre_cfg(struct lustre_cfg *lcfg);
2124 /* quota. fixed by tianzy for bug10707 */
2125 #define QUOTA_IS_GRP 0X1UL /* 0 is user, 1 is group. Used by qd_flags*/
2126 #define QUOTA_IS_BLOCK 0x2UL /* 0 is inode, 1 is block. Used by qd_flags*/
2129 __u32 qd_id; /* ID appiles to (uid, gid) */
2130 __u32 qd_flags; /* Quota type (USRQUOTA, GRPQUOTA) occupy one bit;
2131 * Block quota or file quota occupy one bit */
2132 __u64 qd_count; /* acquire/release count (bytes for block quota) */
2135 struct qunit_data_old {
2136 __u32 qd_id; /* ID appiles to (uid, gid) */
2137 __u32 qd_type; /* Quota type (USRQUOTA, GRPQUOTA) */
2138 __u32 qd_count; /* acquire/release count (bytes for block quota) */
2139 __u32 qd_isblk; /* Block quota or file quota */
2142 extern void lustre_swab_qdata(struct qunit_data *d);
2143 extern void lustre_swab_qdata_old(struct qunit_data_old *d);
2144 extern struct qunit_data *lustre_quota_old_to_new(struct qunit_data_old *d);
2145 extern struct qunit_data_old *lustre_quota_new_to_old(struct qunit_data *d);
2152 #define JOIN_FILE_ALIGN 4096
2154 /* security opcodes */
2157 SEC_CTX_INIT_CONT = 802,
2160 SEC_FIRST_OPC = SEC_CTX_INIT
2164 * capa related definitions
2166 #define CAPA_HMAC_MAX_LEN 64
2167 #define CAPA_HMAC_KEY_MAX_LEN 56
2169 /* NB take care when changing the sequence of elements this struct,
2170 * because the offset info is used in find_capa() */
2171 struct lustre_capa {
2172 struct lu_fid lc_fid; /* fid */
2173 __u64 lc_opc; /* operations allowed */
2174 __u32 lc_uid; /* uid, it is obsolete, but maybe used in
2175 * future, reserve it for 64-bits aligned.*/
2176 __u32 lc_flags; /* HMAC algorithm & flags */
2177 __u32 lc_keyid; /* key used for the capability */
2178 __u32 lc_timeout; /* capa timeout value (sec) */
2179 __u64 lc_expiry; /* expiry time (sec) */
2180 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /* HMAC */
2181 } __attribute__((packed));
2183 extern void lustre_swab_lustre_capa(struct lustre_capa *c);
2185 /* lustre_capa.lc_opc */
2187 CAPA_OPC_BODY_WRITE = 1<<0, /* write object data */
2188 CAPA_OPC_BODY_READ = 1<<1, /* read object data */
2189 CAPA_OPC_INDEX_LOOKUP = 1<<2, /* lookup object fid */
2190 CAPA_OPC_INDEX_INSERT = 1<<3, /* insert object fid */
2191 CAPA_OPC_INDEX_DELETE = 1<<4, /* delete object fid */
2192 CAPA_OPC_OSS_WRITE = 1<<5, /* write oss object data */
2193 CAPA_OPC_OSS_READ = 1<<6, /* read oss object data */
2194 CAPA_OPC_OSS_TRUNC = 1<<7, /* truncate oss object */
2195 CAPA_OPC_META_WRITE = 1<<8, /* write object meta data */
2196 CAPA_OPC_META_READ = 1<<9, /* read object meta data */
2200 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
2201 #define CAPA_OPC_MDS_ONLY \
2202 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
2203 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
2204 #define CAPA_OPC_OSS_ONLY \
2205 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC)
2206 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
2207 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
2209 /* MDS capability covers object capability for operations of body r/w
2210 * (dir readpage/sendpage), index lookup/insert/delete and meta data r/w,
2211 * while OSS capability only covers object capability for operations of
2212 * oss data(file content) r/w/truncate.
2214 static inline int capa_for_mds(struct lustre_capa *c)
2216 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) != 0;
2219 static inline int capa_for_oss(struct lustre_capa *c)
2221 return (c->lc_opc & CAPA_OPC_INDEX_LOOKUP) == 0;
2224 /* lustre_capa.lc_hmac_alg */
2226 CAPA_HMAC_ALG_SHA1 = 1, /* sha1 algorithm */
2230 #define CAPA_FL_MASK 0x00ffffff
2231 #define CAPA_HMAC_ALG_MASK 0xff000000
2233 struct lustre_capa_key {
2234 __u64 lk_mdsid; /* mds# */
2235 __u32 lk_keyid; /* key# */
2237 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /* key */
2238 } __attribute__((packed));
2240 extern void lustre_swab_lustre_capa_key(struct lustre_capa_key *k);
2242 /* quota check function */
2243 #define QUOTA_RET_OK 0 /* return successfully */
2244 #define QUOTA_RET_NOQUOTA 1 /* not support quota */
2245 #define QUOTA_RET_NOLIMIT 2 /* quota limit isn't set */
2246 #define QUOTA_RET_ACQUOTA 3 /* need to acquire extra quota */