4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; If not, see
18 * http://www.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2011, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/lustre/lustre_idl.h
34 * Lustre wire protocol definitions.
37 /** \defgroup lustreidl lustreidl
39 * Lustre wire protocol definitions.
41 * ALL structs passing over the wire should be declared here. Structs
42 * that are used in interfaces with userspace should go in lustre_user.h.
44 * All structs being declared here should be built from simple fixed-size
45 * types (__u8, __u16, __u32, __u64) or be built from other types or
46 * structs also declared in this file. Similarly, all flags and magic
47 * values in those structs should also be declared here. This ensures
48 * that the Lustre wire protocol is not influenced by external dependencies.
50 * The only other acceptable items in this file are VERY SIMPLE accessor
51 * functions to avoid callers grubbing inside the structures. Nothing that
52 * depends on external functions or definitions should be in here.
54 * Structs must be properly aligned to put 64-bit values on an 8-byte
55 * boundary. Any structs being added here must also be added to
56 * utils/wirecheck.c and "make newwiretest" run to regenerate the
57 * utils/wiretest.c sources. This allows us to verify that wire structs
58 * have the proper alignment/size on all architectures.
60 * DO NOT CHANGE any of the structs, flags, values declared here and used
61 * in released Lustre versions. Some structs may have padding fields that
62 * can be used. Some structs might allow addition at the end (verify this
63 * in the code to ensure that new/old clients that see this larger struct
64 * do not fail, otherwise you need to implement protocol compatibility).
69 #ifndef _LUSTRE_IDL_H_
70 #define _LUSTRE_IDL_H_
72 #include <asm/byteorder.h>
73 #include <linux/errno.h>
74 #include <linux/types.h>
77 * This is due to us being out of kernel and the way the OpenSFS branch
81 # include <uapi/linux/lnet/lnet-types.h>
82 # include <uapi/linux/lustre/lustre_user.h> /* Defn's shared with user-space. */
83 # include <uapi/linux/lustre/lustre_ver.h>
85 # include <linux/lnet/lnet-types.h>
86 # include <linux/lustre/lustre_user.h>
87 # include <linux/lustre/lustre_ver.h>
90 #if defined(__cplusplus)
97 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
98 * FOO_REPLY_PORTAL is for incoming replies on the FOO
99 * FOO_BULK_PORTAL is for incoming bulk on the FOO
102 #define CONNMGR_REQUEST_PORTAL 1
103 #define CONNMGR_REPLY_PORTAL 2
104 //#define OSC_REQUEST_PORTAL 3
105 #define OSC_REPLY_PORTAL 4
106 //#define OSC_BULK_PORTAL 5
107 #define OST_IO_PORTAL 6
108 #define OST_CREATE_PORTAL 7
109 #define OST_BULK_PORTAL 8
110 //#define MDC_REQUEST_PORTAL 9
111 #define MDC_REPLY_PORTAL 10
112 //#define MDC_BULK_PORTAL 11
113 #define MDS_REQUEST_PORTAL 12
114 #define MDS_IO_PORTAL 13
115 #define MDS_BULK_PORTAL 14
116 #define LDLM_CB_REQUEST_PORTAL 15
117 #define LDLM_CB_REPLY_PORTAL 16
118 #define LDLM_CANCEL_REQUEST_PORTAL 17
119 #define LDLM_CANCEL_REPLY_PORTAL 18
120 //#define PTLBD_REQUEST_PORTAL 19
121 //#define PTLBD_REPLY_PORTAL 20
122 //#define PTLBD_BULK_PORTAL 21
123 #define MDS_SETATTR_PORTAL 22
124 #define MDS_READPAGE_PORTAL 23
125 #define OUT_PORTAL 24
126 #define MGC_REPLY_PORTAL 25
127 #define MGS_REQUEST_PORTAL 26
128 #define MGS_REPLY_PORTAL 27
129 #define OST_REQUEST_PORTAL 28
130 #define FLD_REQUEST_PORTAL 29
131 #define SEQ_METADATA_PORTAL 30
132 #define SEQ_DATA_PORTAL 31
133 #define SEQ_CONTROLLER_PORTAL 32
134 #define MGS_BULK_PORTAL 33
135 /* #define DVS_PORTAL 63 */
136 /* reserved for Cray DVS - spitzcor@cray.com, roe@cray.com, n8851@cray.com */
139 * Describes a range of sequence, lsr_start is included but lsr_end is
141 * Same structure is used in fld module where lsr_index field holds mdt id
144 struct lu_seq_range {
151 struct lu_seq_range_array {
154 struct lu_seq_range lsra_lsr[0];
157 #define LU_SEQ_RANGE_MDT 0x0
158 #define LU_SEQ_RANGE_OST 0x1
159 #define LU_SEQ_RANGE_ANY 0x3
161 #define LU_SEQ_RANGE_MASK 0x3
163 /** \defgroup lu_fid lu_fid
166 extern void lustre_lma_swab(struct lustre_mdt_attrs *lma);
167 extern void lustre_lma_init(struct lustre_mdt_attrs *lma,
168 const struct lu_fid *fid,
169 __u32 compat, __u32 incompat);
170 extern void lustre_loa_swab(struct lustre_ost_attrs *loa,
172 extern void lustre_loa_init(struct lustre_ost_attrs *loa,
173 const struct lu_fid *fid,
174 __u32 compat, __u32 incompat);
176 /* copytool uses a 32b bitmask field to encode archive-Ids during register
178 * archive num = 0 => all
179 * archive num from 1 to 32
181 #define LL_HSM_MAX_ARCHIVE (sizeof(__u32) * 8)
184 * HSM on-disk attributes stored in a separate xattr.
187 /** Bitfield for supported data in this structure. For future use. */
190 /** HSM flags, see hsm_flags enum below */
192 /** backend archive id associated with the file */
194 /** version associated with the last archiving, if any */
197 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
203 /** LASTID file has zero OID */
204 LUSTRE_FID_LASTID_OID = 0UL,
205 /** initial fid id value */
206 LUSTRE_FID_INIT_OID = 1UL
210 * Different FID Format
211 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
214 * File IDentifier generated by client from range allocated by the seq service.
215 * First 0x400 sequences [2^33, 2^33 + 0x400] are reserved for system use. Note
216 * that on ldiskfs MDTs that IGIF FIDs can use inode numbers starting at 12,
217 * but this is in the IGIF SEQ rangeand does not conflict with assigned FIDs.
220 * Inode and Generation In FID, a surrogate FID used to globally identify an
221 * existing object on OLD formatted MDT file system. This would only be used on
222 * MDT0 in a DNE filesystem, because there are not expected to be any OLD
223 * formatted DNE filesystems. Belongs to a sequence in [12, 2^32 - 1] range,
224 * where sequence number is inode number, and inode generation is used as OID.
225 * NOTE: This assumes no more than 2^32-1 inodes exist in the MDT filesystem,
226 * which is the maximum possible for an ldiskfs backend. NOTE: This assumes
227 * that the reserved ext3/ext4/ldiskfs inode numbers [0-11] are never visible
228 * to clients, which has always been true.
231 * Object ID in FID, a surrogate FID used to globally identify an existing
232 * object on OLD formatted OST file system. Belongs to a sequence in
233 * [2^32, 2^33 - 1]. Sequence number is calculated as:
234 * 1 << 32 | (ost_index << 16) | ((objid >> 32) & 0xffff)
235 * that is, SEQ consists of 16-bit OST index, and higher 16 bits of object ID.
236 * The generation of unique SEQ values per OST allows the IDIF FIDs to be
237 * identified in the FLD correctly. The OID field is calculated as:
239 * that is, it consists of lower 32 bits of object ID. NOTE This assumes that
240 * no more than 2^48-1 objects have ever been created on an OST, and that no
241 * more than 65535 OSTs are in use. Both are very reasonable assumptions (can
242 * uniquely map all objects on an OST that created 1M objects per second for 9
243 * years, or combinations thereof).
246 * Surrogate FID used to identify an existing object on OLD formatted OST
247 * filesystem. Belongs to the reserved sequence 0, and is used internally prior
248 * to the introduction of FID-on-OST, at which point IDIF will be used to
249 * identify objects as residing on a specific OST.
252 * For Lustre Log objects the object sequence 1 is used. This is compatible with
253 * both OLD and NEW.1 namespaces, as this SEQ number is in the ext3/ldiskfs
254 * reserved inode range and does not conflict with IGIF sequence numbers.
257 * For testing OST IO performance the object sequence 2 is used. This is
258 * compatible with both OLD and NEW.1 namespaces, as this SEQ number is in the
259 * ext3/ldiskfs reserved inode range and does not conflict with IGIF sequence
262 * OST_MDT1 .. OST_MAX:
263 * For testing with multiple MDTs the object sequence 3 through 9 is used,
264 * allowing direct mapping of MDTs 1 through 7 respectively, for a total of 8
265 * MDTs including OST_MDT0. This matches the legacy CMD project "group"
266 * mappings. However, this SEQ range is only for testing prior to any production
267 * DNE release, as the objects in this range conflict across all OSTs, as the
268 * OST index is not part of the FID.
271 * For compatibility with existing OLD OST network protocol structures, the FID
272 * must map onto the o_id and o_gr in a manner that ensures existing objects are
273 * identified consistently for IO, as well as onto the lock namespace to ensure
274 * both IDIFs map onto the same objects for IO as well as resources in the DLM.
277 * resource[] = {o_id, o_seq, 0, 0}; // o_seq == 0 for production releases
279 * DLM NEW.1 FID (this is the same for both the MDT and OST):
280 * resource[] = {SEQ, OID, VER, HASH};
282 * Note that for mapping IDIF values to DLM resource names the o_id may be
283 * larger than the 2^33 reserved sequence numbers for IDIF, so it is possible
284 * for the o_id numbers to overlap FID SEQ numbers in the resource. However, in
285 * all production releases the OLD o_seq field is always zero, and all valid FID
286 * OID values are non-zero, so the lock resources will not collide.
288 * For objects within the IDIF range, group extraction (non-CMD) will be:
289 * o_id = (fid->f_seq & 0x7fff) << 16 | fid->f_oid;
290 * o_seq = 0; // formerly group number
294 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
295 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
296 * used for other purposes and not risk collisions with existing inodes.
299 FID_SEQ_OST_MDT0 = 0,
300 FID_SEQ_LLOG = 1, /* unnamed llogs */
302 FID_SEQ_UNUSED_START = 3,
303 FID_SEQ_UNUSED_END = 9,
304 FID_SEQ_LLOG_NAME = 10, /* named llogs */
307 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
308 FID_SEQ_IDIF = 0x100000000ULL,
309 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
310 /* Normal FID sequence starts from this value, i.e. 1<<33 */
311 FID_SEQ_START = 0x200000000ULL,
312 /* sequence for local pre-defined FIDs listed in local_oid */
313 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
314 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
315 /* sequence is used for local named objects FIDs generated
316 * by local_object_storage library */
317 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
318 /* Because current FLD will only cache the fid sequence, instead
319 * of oid on the client side, if the FID needs to be exposed to
320 * clients sides, it needs to make sure all of fids under one
321 * sequence will be located in one MDT. */
322 FID_SEQ_SPECIAL = 0x200000004ULL,
323 FID_SEQ_QUOTA = 0x200000005ULL,
324 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
325 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
326 FID_SEQ_LAYOUT_RBTREE = 0x200000008ULL,
327 /* sequence is used for update logs of cross-MDT operation */
328 FID_SEQ_UPDATE_LOG = 0x200000009ULL,
329 /* Sequence is used for the directory under which update logs
331 FID_SEQ_UPDATE_LOG_DIR = 0x20000000aULL,
332 FID_SEQ_NORMAL = 0x200000400ULL,
333 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
336 #define OBIF_OID_MAX_BITS 32
337 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
338 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
339 #define IDIF_OID_MAX_BITS 48
340 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
341 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
343 /** OID for FID_SEQ_SPECIAL */
345 /* Big Filesystem Lock to serialize rename operations */
346 FID_OID_SPECIAL_BFL = 1UL,
349 /** OID for FID_SEQ_DOT_LUSTRE */
350 enum dot_lustre_oid {
351 FID_OID_DOT_LUSTRE = 1UL,
352 FID_OID_DOT_LUSTRE_OBF = 2UL,
353 FID_OID_DOT_LUSTRE_LPF = 3UL,
356 /** OID for FID_SEQ_ROOT */
359 FID_OID_ECHO_ROOT = 2UL,
362 struct lu_orphan_rec {
363 /* The MDT-object's FID referenced by the orphan OST-object */
364 struct lu_fid lor_fid;
369 struct lu_orphan_ent {
370 /* The orphan OST-object's FID */
371 struct lu_fid loe_key;
372 struct lu_orphan_rec loe_rec;
375 struct lu_orphan_rec_v2 {
376 struct lu_orphan_rec lor_rec;
377 struct ost_layout lor_layout;
381 struct lu_orphan_ent_v2 {
382 /* The orphan OST-object's FID */
383 struct lu_fid loe_key;
384 struct lu_orphan_rec_v2 loe_rec;
389 /** \defgroup lu_dir lu_dir
393 * Enumeration of possible directory entry attributes.
395 * Attributes follow directory entry header in the order they appear in this
398 enum lu_dirent_attrs {
401 LUDA_64BITHASH = 0x0004,
403 /* The following attrs are used for MDT internal only,
404 * not visible to client */
406 /* Something in the record is unknown, to be verified in further. */
407 LUDA_UNKNOWN = 0x0400,
408 /* Ignore this record, go to next directly. */
409 LUDA_IGNORE = 0x0800,
410 /* The system is upgraded, has beed or to be repaired (dryrun). */
411 LUDA_UPGRADE = 0x1000,
412 /* The dirent has been repaired, or to be repaired (dryrun). */
413 LUDA_REPAIR = 0x2000,
414 /* Only check but not repair the dirent inconsistency */
415 LUDA_VERIFY_DRYRUN = 0x4000,
416 /* Verify the dirent consistency */
417 LUDA_VERIFY = 0x8000,
420 #define LU_DIRENT_ATTRS_MASK 0xff00
423 * Layout of readdir pages, as transmitted on wire.
426 /** valid if LUDA_FID is set. */
427 struct lu_fid lde_fid;
428 /** a unique entry identifier: a hash or an offset. */
430 /** total record length, including all attributes. */
434 /** optional variable size attributes following this entry.
435 * taken from enum lu_dirent_attrs.
438 /** name is followed by the attributes indicated in ->ldp_attrs, in
439 * their natural order. After the last attribute, padding bytes are
440 * added to make ->lde_reclen a multiple of 8.
446 * Definitions of optional directory entry attributes formats.
448 * Individual attributes do not have their length encoded in a generic way. It
449 * is assumed that consumer of an attribute knows its format. This means that
450 * it is impossible to skip over an unknown attribute, except by skipping over all
451 * remaining attributes (by using ->lde_reclen), which is not too
452 * constraining, because new server versions will append new attributes at
453 * the end of an entry.
457 * Fid directory attribute: a fid of an object referenced by the entry. This
458 * will be almost always requested by the client and supplied by the server.
460 * Aligned to 8 bytes.
462 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
467 * Aligned to 2 bytes.
474 __u64 ldp_hash_start;
478 struct lu_dirent ldp_entries[0];
481 enum lu_dirpage_flags {
483 * dirpage contains no entry.
487 * last entry's lde_hash equals ldp_hash_end.
492 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
494 if (__le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
497 return dp->ldp_entries;
500 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
502 struct lu_dirent *next;
504 if (__le16_to_cpu(ent->lde_reclen) != 0)
505 next = ((void *)ent) + __le16_to_cpu(ent->lde_reclen);
512 static inline size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
516 if (attr & LUDA_TYPE) {
517 const size_t align = sizeof(struct luda_type) - 1;
518 size = (sizeof(struct lu_dirent) + namelen + align) & ~align;
519 size += sizeof(struct luda_type);
521 size = sizeof(struct lu_dirent) + namelen;
523 return (size + 7) & ~7;
526 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
529 * MDS_READPAGE page size
531 * This is the directory page size packed in MDS_READPAGE RPC.
532 * It's different than PAGE_SIZE because the client needs to
533 * access the struct lu_dirpage header packed at the beginning of
534 * the "page" and without this there isn't any way to know find the
535 * lu_dirpage header is if client and server PAGE_SIZE differ.
537 #define LU_PAGE_SHIFT 12
538 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
539 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
541 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
545 struct lustre_handle {
548 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
550 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
552 return lh->cookie != 0;
555 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
556 const struct lustre_handle *lh2)
558 return lh1->cookie == lh2->cookie;
561 static inline void lustre_handle_copy(struct lustre_handle *tgt,
562 const struct lustre_handle *src)
564 tgt->cookie = src->cookie;
567 struct lustre_handle_array {
569 struct lustre_handle handles[0];
572 /* lustre_msg struct magic. DON'T use swabbed values of MAGIC as magic! */
573 #define LUSTRE_MSG_MAGIC_V2 0x0BD00BD3
574 #define LUSTRE_MSG_MAGIC_V2_SWABBED 0xD30BD00B
575 #define LUSTRE_MSG_MAGIC LUSTRE_MSG_MAGIC_V2
577 /* flags for lm_flags */
578 #define MSGHDR_AT_SUPPORT 0x1 /* adaptive timeouts, lm_cksum valid
579 * in early reply messages */
580 #define MSGHDR_CKSUM_INCOMPAT18 0x2 /* compat for 1.8, needs to be set well
581 * beyond 2.8.0 for compatibility */
582 #define lustre_msg lustre_msg_v2
583 /* we depend on this structure to be 8-byte aligned */
584 /* this type is only endian-adjusted in lustre_unpack_msg() */
585 struct lustre_msg_v2 {
586 __u32 lm_bufcount; /* number of buffers in lm_buflens[] */
587 __u32 lm_secflvr; /* 0 = no crypto, or sptlrpc security flavour */
588 __u32 lm_magic; /* RPC version magic = LUSTRE_MSG_MAGIC_V2 */
589 __u32 lm_repsize; /* size of preallocated reply buffer */
590 __u32 lm_cksum; /* CRC32 of ptlrpc_body early reply messages */
591 __u32 lm_flags; /* MSGHDR_* flags */
592 __u32 lm_padding_2; /* unused */
593 __u32 lm_padding_3; /* unused */
594 __u32 lm_buflens[0]; /* length of additional buffers in bytes,
595 * padded to a multiple of 8 bytes. */
597 * message buffers are packed after padded lm_buflens[] array,
598 * padded to a multiple of 8 bytes each to align contents.
602 /* ptlrpc_body packet pb_types */
603 #define PTL_RPC_MSG_REQUEST 4711 /* normal RPC request message */
604 #define PTL_RPC_MSG_ERR 4712 /* error reply if request unprocessed */
605 #define PTL_RPC_MSG_REPLY 4713 /* normal RPC reply message */
607 /* ptlrpc_body pb_version ((target_version << 16) | rpc_version) */
608 #define PTLRPC_MSG_VERSION 0x00000003
609 #define LUSTRE_VERSION_MASK 0xffff0000
610 #define LUSTRE_OBD_VERSION 0x00010000
611 #define LUSTRE_MDS_VERSION 0x00020000
612 #define LUSTRE_OST_VERSION 0x00030000
613 #define LUSTRE_DLM_VERSION 0x00040000
614 #define LUSTRE_LOG_VERSION 0x00050000
615 #define LUSTRE_MGS_VERSION 0x00060000
617 /* pb_flags that apply to all request messages */
618 /* #define MSG_LAST_REPLAY 0x0001 obsolete 2.0 => {REQ,LOCK}_REPLAY_DONE */
619 #define MSG_RESENT 0x0002 /* was previously sent, no reply seen */
620 #define MSG_REPLAY 0x0004 /* was processed, got reply, recovery */
621 /* #define MSG_AT_SUPPORT 0x0008 obsolete since 1.5, AT always enabled */
622 /* #define MSG_DELAY_REPLAY 0x0010 obsolete since 2.0 */
623 /* #define MSG_VERSION_REPLAY 0x0020 obsolete since 1.8.2, VBR always on */
624 #define MSG_REQ_REPLAY_DONE 0x0040 /* request replay over, locks next */
625 #define MSG_LOCK_REPLAY_DONE 0x0080 /* lock replay over, client done */
627 /* pb_op_flags for connect opcodes: MDS_CONNECT, OST_CONNECT, MGS_CONNECT */
628 #define MSG_CONNECT_RECOVERING 0x00000001 /* target is in recovery */
629 #define MSG_CONNECT_RECONNECT 0x00000002 /* tgt already has client import */
630 #define MSG_CONNECT_REPLAYABLE 0x00000004 /* target supports RPC replay */
631 /* #define MSG_CONNECT_PEER 0x00000008 obsolete since 1.2, removed in 1.5 */
632 #define MSG_CONNECT_LIBCLIENT 0x00000010 /* obsolete since 2.3, removed 2.6 */
633 #define MSG_CONNECT_INITIAL 0x00000020 /* first client connection attempt */
634 /* #define MSG_CONNECT_ASYNC 0x00000040 obsolete since 1.5 */
635 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
636 #define MSG_CONNECT_TRANSNO 0x00000100 /* client sent transno in replay */
638 /* number of previous object versions in pb_pre_versions[] */
639 #define PTLRPC_NUM_VERSIONS 4
640 /* without gss, ptlrpc_body is put at the first buffer. */
641 struct ptlrpc_body_v3 {
642 struct lustre_handle pb_handle;
643 __u32 pb_type; /* request/reply/err type: PTL_RPC_MSG_* */
644 __u32 pb_version; /* LUSTRE_*_VERSION | PTLRPC_MSG_VERSION */
645 __u32 pb_opc; /* RPC opcodes: MDS_*, OST_*, LDLM_, ... */
646 __u32 pb_status; /* negative Linux x86 error number */
647 __u64 pb_last_xid; /* highest replied XID w/o lower unreplied XID*/
648 __u16 pb_tag; /* multiple modifying RPCs virtual slot index */
651 __u64 pb_last_committed;/* rep: highest pb_transno committed to disk */
652 __u64 pb_transno; /* server-assigned transno for modifying RPCs */
653 __u32 pb_flags; /* req: MSG_* flags */
654 __u32 pb_op_flags; /* req: MSG_CONNECT_* flags */
655 __u32 pb_conn_cnt; /* connect instance of this client on server */
656 __u32 pb_timeout; /* req: max wait time; rep: service estimate */
657 __u32 pb_service_time; /* rep: server arrival to reply in seconds */
658 __u32 pb_limit; /* rep: dynamic DLM LRU lock count limit */
659 __u64 pb_slv; /* rep: dynamic DLM LRU server lock volume */
660 /* VBR: rep: previous pb_version(s) of objects modified by this RPC */
661 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
662 __u64 pb_mbits; /**< match bits for bulk request */
663 /* padding for future needs - fix lustre_swab_ptlrpc_body() also */
664 __u64 pb_padding64_0;
665 __u64 pb_padding64_1;
666 __u64 pb_padding64_2;
667 char pb_jobid[LUSTRE_JOBID_SIZE]; /* req: ASCII MPI jobid from env */
669 #define ptlrpc_body ptlrpc_body_v3
671 struct ptlrpc_body_v2 {
672 struct lustre_handle pb_handle;
677 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
678 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
681 __u64 pb_last_committed;
686 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
687 __u32 pb_service_time; /* for rep, actual service time, also used for
688 net_latency of req */
691 /* VBR: pre-versions */
692 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
693 __u64 pb_mbits; /**< unused in V2 */
694 /* padding for future needs */
695 __u64 pb_padding64_0;
696 __u64 pb_padding64_1;
697 __u64 pb_padding64_2;
700 /* message body offset for lustre_msg_v2 */
701 /* ptlrpc body offset in all request/reply messages */
702 #define MSG_PTLRPC_BODY_OFF 0
704 /* normal request/reply message record offset */
705 #define REQ_REC_OFF 1
706 #define REPLY_REC_OFF 1
708 /* ldlm request message body offset */
709 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
710 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
712 /* ldlm intent lock message body offset */
713 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
714 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
716 /* ldlm reply message body offset */
717 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
718 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
720 /** only use in req->rq_{req,rep}_swab_mask */
721 #define MSG_PTLRPC_HEADER_OFF 31
724 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
725 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
726 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
727 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
728 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
729 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
730 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
731 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
732 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
733 #define OBD_CONNECT_LARGE_ACL 0x200ULL /* more than 32 ACL entries */
734 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
735 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
736 #define OBD_CONNECT_IBITS 0x1000ULL /* not checked in 2.11+ */
737 #define OBD_CONNECT_BARRIER 0x2000ULL /* write barrier */
738 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
739 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
740 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
741 * in production. Removed in
742 * 2.9. Keep this flag to
745 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
746 * never used in production.
747 * Removed in 2.9. Keep this
748 * flag to avoid reusing.
750 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
751 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
752 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
753 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
754 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
755 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
756 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
757 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
758 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
759 #define OBD_CONNECT_REAL 0x8000000ULL /* obsolete since 2.8 */
760 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
761 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
762 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
763 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
764 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
765 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
766 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
767 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
768 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
769 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
770 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
772 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
773 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
774 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
775 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
776 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
777 * RPC error properly */
778 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
779 * finer space reservation */
780 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
781 * policy and 2.x server */
782 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
783 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
784 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
785 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
786 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
787 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* improved flock deadlock detection */
788 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/* create stripe disposition*/
789 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
791 #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */
792 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */
793 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify
795 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
796 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
797 #define OBD_CONNECT_LOCKAHEAD_OLD 0x1000000000000000ULL /* Old Cray lockahead */
799 /** bulk matchbits is sent within ptlrpc_body */
800 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL
801 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
802 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
803 /* ocd_connect_flags2 flags */
804 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* set file security context at create */
805 #define OBD_CONNECT2_LOCKAHEAD 0x2ULL /* ladvise lockahead v2 */
808 * Please DO NOT add flag values here before first ensuring that this same
809 * flag value is not in use on some other branch. Please clear any such
810 * changes with senior engineers before starting to use a new flag. Then,
811 * submit a small patch against EVERY branch that ONLY adds the new flag,
812 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
813 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
814 * can be approved and landed easily to reserve the flag for future use. */
816 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
817 * connection. It is a temporary bug fix for Imperative Recovery interop
818 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
819 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
820 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
822 #define OCD_HAS_FLAG(ocd, flg) \
823 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
826 #ifdef HAVE_LRU_RESIZE_SUPPORT
827 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
829 #define LRU_RESIZE_CONNECT_FLAG 0
832 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
833 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
834 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
835 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
836 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
837 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
838 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
839 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
840 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
841 OBD_CONNECT_EINPROGRESS | \
842 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
843 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
844 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
845 OBD_CONNECT_FLOCK_DEAD | \
846 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
847 OBD_CONNECT_OPEN_BY_FID | \
848 OBD_CONNECT_DIR_STRIPE | OBD_CONNECT_GRANT | \
849 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_SRVLOCK | \
850 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_CKSUM | \
851 OBD_CONNECT_MULTIMODRPCS | \
852 OBD_CONNECT_SUBTREE | OBD_CONNECT_LARGE_ACL | \
853 OBD_CONNECT_GRANT_PARAM | \
854 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
856 #define MDT_CONNECT_SUPPORTED2 OBD_CONNECT2_FILE_SECCTX
858 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
859 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
860 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
861 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
862 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
863 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
864 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
865 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
866 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
867 OBD_CONNECT_MAX_EASIZE | \
868 OBD_CONNECT_EINPROGRESS | \
869 OBD_CONNECT_JOBSTATS | \
870 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
871 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
872 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
873 OBD_CONNECT_BULK_MBITS | \
874 OBD_CONNECT_GRANT_PARAM | \
875 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
877 #define OST_CONNECT_SUPPORTED2 OBD_CONNECT2_LOCKAHEAD
879 #define ECHO_CONNECT_SUPPORTED 0
880 #define ECHO_CONNECT_SUPPORTED2 0
882 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
883 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
884 OBD_CONNECT_MNE_SWAB | OBD_CONNECT_PINGLESS |\
885 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_BARRIER)
887 #define MGS_CONNECT_SUPPORTED2 0
889 /* Features required for this version of the client to work with server */
890 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_FID | \
891 OBD_CONNECT_ATTRFID | \
894 /* This structure is used for both request and reply.
896 * If we eventually have separate connect data for different types, which we
897 * almost certainly will, then perhaps we stick a union in here. */
898 struct obd_connect_data {
899 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
900 __u32 ocd_version; /* lustre release version number */
901 __u32 ocd_grant; /* initial cache grant amount (bytes) */
902 __u32 ocd_index; /* LOV index to connect to */
903 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
904 __u64 ocd_ibits_known; /* inode bits this client understands */
905 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
906 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
907 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
908 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
909 __u64 ocd_transno; /* first transno from client to be replayed */
910 __u32 ocd_group; /* MDS group on OST */
911 __u32 ocd_cksum_types; /* supported checksum algorithms */
912 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
913 __u32 ocd_instance; /* instance # of this target */
914 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
915 /* Fields after ocd_maxbytes are only accessible by the receiver
916 * if the corresponding flag in ocd_connect_flags is set. Accessing
917 * any field after ocd_maxbytes on the receiver without a valid flag
918 * may result in out-of-bound memory access and kernel oops. */
919 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
920 __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */
921 __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */
922 __u64 ocd_connect_flags2;
923 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
924 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
925 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
926 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
927 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
928 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
929 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
930 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
931 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
932 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
933 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
934 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
935 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
938 * Please DO NOT use any fields here before first ensuring that this same
939 * field is not in use on some other branch. Please clear any such changes
940 * with senior engineers before starting to use a new field. Then, submit
941 * a small patch against EVERY branch that ONLY adds the new field along with
942 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
943 * reserve the flag for future use. */
946 * Supported checksum algorithms. Up to 32 checksum types are supported.
947 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
948 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
949 * algorithm and also the OBD_FL_CKSUM* flags.
951 typedef enum cksum_types {
952 OBD_CKSUM_CRC32 = 0x00000001,
953 OBD_CKSUM_ADLER = 0x00000002,
954 OBD_CKSUM_CRC32C= 0x00000004,
958 * OST requests: OBDO & OBD request records
963 OST_REPLY = 0, /* reply ? */
979 OST_QUOTACHECK = 18, /* not used since 2.4 */
981 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
983 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
985 #define OST_FIRST_OPC OST_REPLY
988 OBD_FL_INLINEDATA = 0x00000001,
989 OBD_FL_OBDMDEXISTS = 0x00000002,
990 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
991 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
992 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
993 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
994 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
995 OBD_FL_NO_PRJQUOTA = 0x00000080, /* the object's project is over
997 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
998 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
999 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1000 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1001 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1002 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1003 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1004 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
1005 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1006 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1007 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1008 * XXX: obsoleted - reserved for old
1009 * clients prior than 2.2 */
1010 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1011 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1012 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1013 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1014 /* OBD_FL_LOCAL_MASK = 0xF0000000, was local-only flags until 2.10 */
1016 /* Note that while these checksum values are currently separate bits,
1017 * in 2.x we can actually allow all values from 1-31 if we wanted. */
1018 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1019 OBD_FL_CKSUM_CRC32C,
1023 * All LOV EA magics should have the same postfix, if some new version
1024 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1025 * Lustre, even though the old version system does not recognizes such
1026 * new magic, it still can distinguish the corrupted cases by checking
1027 * the magic's postfix.
1029 #define LOV_MAGIC_MAGIC 0x0BD0
1030 #define LOV_MAGIC_MASK 0xFFFF
1032 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1033 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1034 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1035 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1036 /* reserved for specifying OSTs */
1037 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1038 #define LOV_MAGIC LOV_MAGIC_V1
1039 #define LOV_MAGIC_COMP_V1 (0x0BD60000 | LOV_MAGIC_MAGIC)
1042 * magic for fully defined striping
1043 * the idea is that we should have different magics for striping "hints"
1044 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1045 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1046 * we can't just change it w/o long way preparation, but we still need a
1047 * mechanism to allow LOD to differentiate hint versus ready striping.
1048 * so, at the moment we do a trick: MDT knows what to expect from request
1049 * depending on the case (replay uses ready striping, non-replay req uses
1050 * hints), so MDT replaces magic with appropriate one and now LOD can
1051 * easily understand what's inside -bzzz
1053 * those *_DEF magics are only used on server side internally, they
1054 * won't be put on wire or disk.
1056 #define LOV_MAGIC_DEFINED 0x10000000
1057 #define LOV_MAGIC_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V1)
1058 #define LOV_MAGIC_V3_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V3)
1059 #define LOV_MAGIC_COMP_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_COMP_V1)
1061 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1062 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1064 #define lov_ost_data lov_ost_data_v1
1065 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1066 struct ost_id l_ost_oi; /* OST object ID */
1067 __u32 l_ost_gen; /* generation of this l_ost_idx */
1068 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1071 #define lov_mds_md lov_mds_md_v1
1072 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1073 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1074 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1075 struct ost_id lmm_oi; /* LOV object ID */
1076 __u32 lmm_stripe_size; /* size of stripe in bytes */
1077 /* lmm_stripe_count used to be __u32 */
1078 __u16 lmm_stripe_count; /* num stripes in use for this object */
1079 __u16 lmm_layout_gen; /* layout generation number */
1080 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1083 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1084 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1086 /* This is the default MDT reply size allocated, should the striping be bigger,
1087 * it will be reallocated in mdt_fix_reply.
1088 * 100 stripes is a bit less than 2.5k of data */
1089 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1090 100 * sizeof(struct lov_ost_data))
1092 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1093 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1094 #define XATTR_USER_PREFIX "user."
1095 #define XATTR_TRUSTED_PREFIX "trusted."
1096 #define XATTR_SECURITY_PREFIX "security."
1098 #define XATTR_NAME_SOM "trusted.som"
1099 #define XATTR_NAME_LOV "trusted.lov"
1100 #define XATTR_NAME_LMA "trusted.lma"
1101 #define XATTR_NAME_LMV "trusted.lmv"
1102 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1103 #define XATTR_NAME_LINK "trusted.link"
1104 #define XATTR_NAME_FID "trusted.fid"
1105 #define XATTR_NAME_VERSION "trusted.version"
1106 #define XATTR_NAME_SOM "trusted.som"
1107 #define XATTR_NAME_HSM "trusted.hsm"
1108 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1109 #define XATTR_NAME_DUMMY "trusted.dummy"
1111 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1112 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1114 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1115 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1116 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1117 struct ost_id lmm_oi; /* LOV object ID */
1118 __u32 lmm_stripe_size; /* size of stripe in bytes */
1119 /* lmm_stripe_count used to be __u32 */
1120 __u16 lmm_stripe_count; /* num stripes in use for this object */
1121 __u16 lmm_layout_gen; /* layout generation number */
1122 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1123 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1126 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1128 if (stripes == (__u16)-1)
1131 if (lmm_magic == LOV_MAGIC_V3)
1132 return sizeof(struct lov_mds_md_v3) +
1133 stripes * sizeof(struct lov_ost_data_v1);
1135 return sizeof(struct lov_mds_md_v1) +
1136 stripes * sizeof(struct lov_ost_data_v1);
1140 lov_mds_md_max_stripe_count(size_t buf_size, __u32 lmm_magic)
1142 switch (lmm_magic) {
1143 case LOV_MAGIC_V1: {
1144 struct lov_mds_md_v1 lmm;
1146 if (buf_size < sizeof(lmm))
1149 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1151 case LOV_MAGIC_V3: {
1152 struct lov_mds_md_v3 lmm;
1154 if (buf_size < sizeof(lmm))
1157 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1164 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1165 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1166 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1167 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1168 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1169 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1170 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1171 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1172 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1173 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1174 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1175 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1176 #define OBD_MD_DOM_SIZE (0X00001000ULL) /* Data-on-MDT component size */
1177 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1178 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1179 #define OBD_MD_LAYOUT_VERSION (0x00008000ULL) /* layout version for
1181 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1182 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1183 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1184 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1185 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1186 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1187 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1188 #define OBD_MD_FLPRJQUOTA (0x00400000ULL) /* over quota flags sent from ost */
1189 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1190 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1191 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1192 /* ->mds if epoch opens or closes */
1193 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1194 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1195 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1196 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1197 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1199 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1200 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1201 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1202 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1204 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1205 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1206 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1207 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1208 /* OBD_MD_FLRMTPERM (0x0000010000000000ULL) remote perm, obsolete */
1209 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1210 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1211 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1212 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1213 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1214 * under lock; for xattr
1215 * requests means the
1216 * client holds the lock */
1217 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1219 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1220 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1223 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1224 #define OBD_MD_FLOSTLAYOUT (0x0080000000000000ULL) /* contain ost_layout */
1225 #define OBD_MD_FLPROJID (0x0100000000000000ULL) /* project ID */
1227 #define OBD_MD_FLALLQUOTA (OBD_MD_FLUSRQUOTA | \
1228 OBD_MD_FLGRPQUOTA | \
1231 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1232 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1233 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1234 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1235 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP | \
1238 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1240 /* don't forget obdo_fid which is way down at the bottom so it can
1241 * come after the definition of llog_cookie */
1245 HSS_CLEARMASK = 0x02,
1246 HSS_ARCHIVE_ID = 0x04,
1249 struct hsm_state_set {
1251 __u32 hss_archive_id;
1253 __u64 hss_clearmask;
1256 /* ost_body.data values for OST_BRW */
1258 #define OBD_BRW_READ 0x01
1259 #define OBD_BRW_WRITE 0x02
1260 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1261 #define OBD_BRW_NDELAY 0x04 /* Non-delay RPC should be issued for
1262 * this page. Non-delay RPCs have bit
1263 * rq_no_delay set. */
1264 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1265 * transfer and is not accounted in
1267 #define OBD_BRW_CHECK 0x10
1268 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1269 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1270 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1271 #define OBD_BRW_NOQUOTA 0x100
1272 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1273 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1274 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1275 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1276 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1277 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1278 * that the client is running low on
1279 * space for unstable pages; asking
1280 * it to sync quickly */
1281 #define OBD_BRW_OVER_PRJQUOTA 0x8000 /* Running out of project quota */
1283 #define OBD_BRW_OVER_ALLQUOTA (OBD_BRW_OVER_USRQUOTA | \
1284 OBD_BRW_OVER_GRPQUOTA | \
1285 OBD_BRW_OVER_PRJQUOTA)
1287 #define OBD_BRW_LOCAL1 0x80000000UL /*
1288 * osd-ldiskfs internal,
1289 * page mapped to real block
1292 #define OBD_BRW_LOCALS (OBD_BRW_LOCAL1)
1294 #define OBD_OBJECT_EOF LUSTRE_EOF
1296 #define OST_MIN_PRECREATE 32
1297 #define OST_MAX_PRECREATE 20000
1300 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1301 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1302 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1303 * high 16 bits in 2.4 and later */
1304 __u32 ioo_bufcnt; /* number of niobufs for this object */
1307 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1308 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1309 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. */
1310 #define IOOBJ_MAX_BRW_BITS 16
1311 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1312 #define ioobj_max_brw_set(ioo, num) \
1313 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1315 /* multiple of 8 bytes => can array */
1316 struct niobuf_remote {
1322 /* lock value block communicated between the filter and llite */
1324 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1325 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1326 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1327 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1328 #define OST_LVB_IS_ERR(blocks) \
1329 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1330 #define OST_LVB_SET_ERR(blocks, rc) \
1331 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1332 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1355 * lquota data structures
1358 #ifndef QUOTABLOCK_BITS
1359 # define QUOTABLOCK_BITS LUSTRE_QUOTABLOCK_BITS
1362 #ifndef QUOTABLOCK_SIZE
1363 # define QUOTABLOCK_SIZE LUSTRE_QUOTABLOCK_SIZE
1367 # define toqb lustre_stoqb
1370 /* The lquota_id structure is an union of all the possible identifier types that
1371 * can be used with quota, this includes:
1374 * - a FID which can be used for per-directory quota in the future */
1376 struct lu_fid qid_fid; /* FID for per-directory quota */
1377 __u64 qid_uid; /* user identifier */
1378 __u64 qid_gid; /* group identifier */
1379 __u64 qid_projid; /* project identifier */
1382 /* quotactl management */
1383 struct obd_quotactl {
1385 __u32 qc_type; /* see Q_* flag below */
1388 struct obd_dqinfo qc_dqinfo;
1389 struct obd_dqblk qc_dqblk;
1392 #define Q_COPY(out, in, member) (out)->member = (in)->member
1394 #define QCTL_COPY(out, in) \
1396 Q_COPY(out, in, qc_cmd); \
1397 Q_COPY(out, in, qc_type); \
1398 Q_COPY(out, in, qc_id); \
1399 Q_COPY(out, in, qc_stat); \
1400 Q_COPY(out, in, qc_dqinfo); \
1401 Q_COPY(out, in, qc_dqblk); \
1404 /* Body of quota request used for quota acquire/release RPCs between quota
1405 * master (aka QMT) and slaves (ak QSD). */
1407 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1408 * and type (data or metadata) as well as
1409 * the quota type (user or group). */
1410 union lquota_id qb_id; /* uid or gid or directory FID */
1411 __u32 qb_flags; /* see below */
1413 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1414 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1415 __u64 qb_slv_ver; /* slave index file version */
1416 struct lustre_handle qb_lockh; /* per-ID lock handle */
1417 struct lustre_handle qb_glb_lockh; /* global lock handle */
1418 __u64 qb_padding1[4];
1421 /* When the quota_body is used in the reply of quota global intent
1422 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1423 #define qb_slv_fid qb_fid
1424 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1426 #define qb_qunit qb_usage
1428 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1429 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1430 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1431 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1433 /* Quota types currently supported */
1435 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1436 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1437 LQUOTA_TYPE_PRJ = 0x02, /* maps to PRJQUOTA */
1441 /* There are 2 different resource types on which a quota limit can be enforced:
1442 * - inodes on the MDTs
1443 * - blocks on the OSTs */
1445 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1446 LQUOTA_RES_DT = 0x02,
1448 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1450 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1453 * Space accounting support
1454 * Format of an accounting record, providing disk usage information for a given
1457 struct lquota_acct_rec { /* 16 bytes */
1458 __u64 bspace; /* current space in use */
1459 __u64 ispace; /* current # inodes in use */
1463 * Global quota index support
1464 * Format of a global record, providing global quota settings for a given quota
1467 struct lquota_glb_rec { /* 32 bytes */
1468 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1469 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1470 __u64 qbr_time; /* grace time, in seconds */
1471 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1476 * Slave index support
1477 * Format of a slave record, recording how much space is granted to a given
1480 struct lquota_slv_rec { /* 8 bytes */
1481 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1482 * in #inodes or kbytes */
1485 /* Data structures associated with the quota locks */
1487 /* Glimpse descriptor used for the index & per-ID quota locks */
1488 struct ldlm_gl_lquota_desc {
1489 union lquota_id gl_id; /* quota ID subject to the glimpse */
1490 __u64 gl_flags; /* see LQUOTA_FL* below */
1491 __u64 gl_ver; /* new index version */
1492 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1493 __u64 gl_softlimit; /* new softlimit */
1497 #define gl_qunit gl_hardlimit /* current qunit value used when
1498 * glimpsing per-ID quota locks */
1500 /* quota glimpse flags */
1501 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1503 /* LVB used with quota (global and per-ID) locks */
1505 __u64 lvb_flags; /* see LQUOTA_FL* above */
1506 __u64 lvb_id_may_rel; /* space that might be released later */
1507 __u64 lvb_id_rel; /* space released by the slave for this ID */
1508 __u64 lvb_id_qunit; /* current qunit value */
1512 /* LVB used with global quota lock */
1513 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1521 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1530 MDS_GETATTR_NAME = 34,
1535 MDS_DISCONNECT = 39,
1538 MDS_PIN = 42, /* obsolete, never used in a release */
1539 MDS_UNPIN = 43, /* obsolete, never used in a release */
1541 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1543 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1546 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1548 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1550 MDS_HSM_STATE_GET = 54,
1551 MDS_HSM_STATE_SET = 55,
1552 MDS_HSM_ACTION = 56,
1553 MDS_HSM_PROGRESS = 57,
1554 MDS_HSM_REQUEST = 58,
1555 MDS_HSM_CT_REGISTER = 59,
1556 MDS_HSM_CT_UNREGISTER = 60,
1557 MDS_SWAP_LAYOUTS = 61,
1561 #define MDS_FIRST_OPC MDS_GETATTR
1564 /* opcodes for object update */
1570 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1590 /* the disposition of the intent outlines what was executed */
1591 #define DISP_IT_EXECD 0x00000001
1592 #define DISP_LOOKUP_EXECD 0x00000002
1593 #define DISP_LOOKUP_NEG 0x00000004
1594 #define DISP_LOOKUP_POS 0x00000008
1595 #define DISP_OPEN_CREATE 0x00000010
1596 #define DISP_OPEN_OPEN 0x00000020
1597 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1598 #define DISP_ENQ_OPEN_REF 0x00800000
1599 #define DISP_ENQ_CREATE_REF 0x01000000
1600 #define DISP_OPEN_LOCK 0x02000000
1601 #define DISP_OPEN_LEASE 0x04000000
1602 #define DISP_OPEN_STRIPE 0x08000000
1603 #define DISP_OPEN_DENY 0x10000000
1605 /* INODE LOCK PARTS */
1606 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
1607 * was used to protect permission (mode,
1608 * owner, group etc) before 2.4. */
1609 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1610 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1611 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1613 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
1614 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
1615 * Because for remote directories(in DNE), these locks will be granted by
1616 * different MDTs(different ldlm namespace).
1618 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
1619 * For Remote directory, the master MDT, where the remote directory is, will
1620 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
1621 * will grant LOOKUP_LOCK. */
1622 #define MDS_INODELOCK_PERM 0x000010
1623 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
1624 #define MDS_INODELOCK_DOM 0x000040 /* Data for data-on-mdt files */
1626 #define MDS_INODELOCK_MAXSHIFT 6
1627 /* This FULL lock is useful to take on unlink sort of operations */
1628 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1629 /* DOM lock shouldn't be canceled early, use this macro for ELC */
1630 #define MDS_INODELOCK_ELC (MDS_INODELOCK_FULL & ~MDS_INODELOCK_DOM)
1632 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1633 * but was moved into name[1] along with the OID to avoid consuming the
1634 * name[2,3] fields that need to be used for the quota id (also a FID). */
1636 LUSTRE_RES_ID_SEQ_OFF = 0,
1637 LUSTRE_RES_ID_VER_OID_OFF = 1,
1638 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1639 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1640 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1641 LUSTRE_RES_ID_HSH_OFF = 3
1644 #define MDS_STATUS_CONN 1
1645 #define MDS_STATUS_LOV 2
1648 /* these should be identical to their EXT4_*_FL counterparts, they are
1649 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1650 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1651 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1652 LUSTRE_APPEND_FL = 0x00000020, /* writes to file may only append */
1653 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1654 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1655 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1656 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1657 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1658 LUSTRE_DIRECTIO_FL = 0x00100000, /* Use direct i/o */
1659 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1660 LUSTRE_PROJINHERIT_FL = 0x20000000, /* Create with parents projid */
1662 /* These flags will not be identical to any EXT4_*_FL counterparts,
1663 * and only reserved for lustre purpose. Note: these flags might
1664 * be conflict with some of EXT4 flags, so
1665 * 1. these conflict flags needs to be removed when the flag is
1666 * wired by la_flags see osd_attr_get().
1667 * 2. If these flags needs to be stored into inode, they will be
1668 * stored in LMA. see LMAI_XXXX */
1669 LUSTRE_ORPHAN_FL = 0x00002000,
1671 LUSTRE_LMA_FL_MASKS = LUSTRE_ORPHAN_FL,
1674 #ifndef FS_XFLAG_PROJINHERIT
1675 #define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
1679 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1680 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1681 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1682 * the S_* flags are kernel-internal values that change between kernel
1683 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1684 * See b=16526 for a full history. */
1685 static inline int ll_ext_to_inode_flags(int flags)
1687 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1688 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1689 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1690 #if defined(S_DIRSYNC)
1691 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1693 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0) |
1694 ((flags & LUSTRE_PROJINHERIT_FL) ? FS_XFLAG_PROJINHERIT : 0));
1697 static inline int ll_inode_to_ext_flags(int iflags)
1699 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1700 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1701 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1702 #if defined(S_DIRSYNC)
1703 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1705 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0) |
1706 ((iflags & FS_XFLAG_PROJINHERIT) ? LUSTRE_PROJINHERIT_FL : 0));
1710 /* 64 possible states */
1711 enum md_transient_state {
1712 MS_RESTORE = (1 << 0), /* restore is running */
1716 struct lu_fid mbo_fid1;
1717 struct lu_fid mbo_fid2;
1718 struct lustre_handle mbo_handle;
1720 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1724 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1726 __u64 mbo_t_state; /* transient file state defined in
1727 * enum md_transient_state
1728 * was "ino" until 2.4.0 */
1731 __u32 mbo_capability;
1735 __u32 mbo_flags; /* LUSTRE_*_FL file attributes */
1737 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1738 __u32 mbo_unused2; /* was "generation" until 2.4.0 */
1740 __u32 mbo_eadatasize;
1742 __u32 mbo_max_mdsize;
1743 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1744 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1745 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1747 __u64 mbo_dom_size; /* size of DOM component */
1748 __u64 mbo_dom_blocks; /* blocks consumed by DOM component */
1749 __u64 mbo_padding_8; /* also fix lustre_swab_mdt_body */
1750 __u64 mbo_padding_9;
1751 __u64 mbo_padding_10;
1754 struct mdt_ioepoch {
1755 struct lustre_handle mio_handle;
1756 __u64 mio_unused1; /* was ioepoch */
1757 __u32 mio_unused2; /* was flags */
1761 /* permissions for md_perm.mp_perm */
1763 CFS_SETUID_PERM = 0x01,
1764 CFS_SETGID_PERM = 0x02,
1765 CFS_SETGRP_PERM = 0x04,
1768 struct mdt_rec_setattr {
1778 __u32 sa_padding_1_h;
1779 struct lu_fid sa_fid;
1788 __u32 sa_attr_flags;
1790 __u32 sa_bias; /* some operation flags */
1797 * Attribute flags used in mdt_rec_setattr::sa_valid.
1798 * The kernel's #defines for ATTR_* should not be used over the network
1799 * since the client and MDS may run different kernels (see bug 13828)
1800 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1802 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1803 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1804 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1805 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1806 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1807 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1808 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1809 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1810 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1811 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1812 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1813 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1814 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1815 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1816 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1817 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1818 #define MDS_ATTR_PROJID 0x10000ULL /* = 65536 */
1821 #define FMODE_READ 00000001
1822 #define FMODE_WRITE 00000002
1825 #define MDS_FMODE_CLOSED 00000000
1826 #define MDS_FMODE_EXEC 00000004
1827 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1828 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1829 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1831 #define MDS_OPEN_CREATED 00000010
1832 #define MDS_OPEN_CROSS 00000020
1834 #define MDS_OPEN_CREAT 00000100
1835 #define MDS_OPEN_EXCL 00000200
1836 #define MDS_OPEN_TRUNC 00001000
1837 #define MDS_OPEN_APPEND 00002000
1838 #define MDS_OPEN_SYNC 00010000
1839 #define MDS_OPEN_DIRECTORY 00200000
1841 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1842 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1843 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1844 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1845 * We do not support JOIN FILE
1846 * anymore, reserve this flags
1847 * just for preventing such bit
1850 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1851 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1852 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1853 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1854 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1856 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1858 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1859 * delegation, succeed if it's not
1860 * being opened with conflict mode.
1862 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1864 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1866 /* lustre internal open flags, which should not be set from user space */
1867 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1868 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1869 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1870 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC)
1873 MDS_CHECK_SPLIT = 1 << 0,
1874 MDS_CROSS_REF = 1 << 1,
1875 MDS_VTX_BYPASS = 1 << 2,
1876 MDS_PERM_BYPASS = 1 << 3,
1877 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1878 MDS_QUOTA_IGNORE = 1 << 5,
1879 /* Was MDS_CLOSE_CLEANUP (1 << 6), No more used */
1880 MDS_KEEP_ORPHAN = 1 << 7,
1881 MDS_RECOV_OPEN = 1 << 8,
1882 MDS_DATA_MODIFIED = 1 << 9,
1883 MDS_CREATE_VOLATILE = 1 << 10,
1884 MDS_OWNEROVERRIDE = 1 << 11,
1885 MDS_HSM_RELEASE = 1 << 12,
1886 MDS_RENAME_MIGRATE = 1 << 13,
1887 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1888 MDS_CLOSE_LAYOUT_MERGE = 1 << 15,
1889 MDS_CLOSE_RESYNC_DONE = 1 << 16,
1892 #define MDS_CLOSE_INTENT (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP | \
1893 MDS_CLOSE_LAYOUT_MERGE | MDS_CLOSE_RESYNC_DONE)
1895 /* instance of mdt_reint_rec */
1896 struct mdt_rec_create {
1904 __u32 cr_suppgid1_h;
1906 __u32 cr_suppgid2_h;
1907 struct lu_fid cr_fid1;
1908 struct lu_fid cr_fid2;
1909 struct lustre_handle cr_old_handle; /* handle in case of open replay */
1913 __u64 cr_padding_1; /* rr_blocks */
1916 /* use of helpers set/get_mrc_cr_flags() is needed to access
1917 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
1918 * extend cr_flags size without breaking 1.8 compat */
1919 __u32 cr_flags_l; /* for use with open, low 32 bits */
1920 __u32 cr_flags_h; /* for use with open, high 32 bits */
1921 __u32 cr_umask; /* umask for create */
1922 __u32 cr_padding_4; /* rr_padding_4 */
1925 /* instance of mdt_reint_rec */
1926 struct mdt_rec_link {
1934 __u32 lk_suppgid1_h;
1936 __u32 lk_suppgid2_h;
1937 struct lu_fid lk_fid1;
1938 struct lu_fid lk_fid2;
1940 __u64 lk_padding_1; /* rr_atime */
1941 __u64 lk_padding_2; /* rr_ctime */
1942 __u64 lk_padding_3; /* rr_size */
1943 __u64 lk_padding_4; /* rr_blocks */
1945 __u32 lk_padding_5; /* rr_mode */
1946 __u32 lk_padding_6; /* rr_flags */
1947 __u32 lk_padding_7; /* rr_padding_2 */
1948 __u32 lk_padding_8; /* rr_padding_3 */
1949 __u32 lk_padding_9; /* rr_padding_4 */
1952 /* instance of mdt_reint_rec */
1953 struct mdt_rec_unlink {
1961 __u32 ul_suppgid1_h;
1963 __u32 ul_suppgid2_h;
1964 struct lu_fid ul_fid1;
1965 struct lu_fid ul_fid2;
1967 __u64 ul_padding_2; /* rr_atime */
1968 __u64 ul_padding_3; /* rr_ctime */
1969 __u64 ul_padding_4; /* rr_size */
1970 __u64 ul_padding_5; /* rr_blocks */
1973 __u32 ul_padding_6; /* rr_flags */
1974 __u32 ul_padding_7; /* rr_padding_2 */
1975 __u32 ul_padding_8; /* rr_padding_3 */
1976 __u32 ul_padding_9; /* rr_padding_4 */
1979 /* instance of mdt_reint_rec */
1980 struct mdt_rec_rename {
1988 __u32 rn_suppgid1_h;
1990 __u32 rn_suppgid2_h;
1991 struct lu_fid rn_fid1;
1992 struct lu_fid rn_fid2;
1994 __u64 rn_padding_1; /* rr_atime */
1995 __u64 rn_padding_2; /* rr_ctime */
1996 __u64 rn_padding_3; /* rr_size */
1997 __u64 rn_padding_4; /* rr_blocks */
1998 __u32 rn_bias; /* some operation flags */
1999 __u32 rn_mode; /* cross-ref rename has mode */
2000 __u32 rn_padding_5; /* rr_flags */
2001 __u32 rn_padding_6; /* rr_padding_2 */
2002 __u32 rn_padding_7; /* rr_padding_3 */
2003 __u32 rn_padding_8; /* rr_padding_4 */
2006 /* instance of mdt_reint_rec */
2007 struct mdt_rec_setxattr {
2015 __u32 sx_suppgid1_h;
2017 __u32 sx_suppgid2_h;
2018 struct lu_fid sx_fid;
2019 __u64 sx_padding_1; /* These three are rr_fid2 */
2024 __u64 sx_padding_5; /* rr_ctime */
2025 __u64 sx_padding_6; /* rr_size */
2026 __u64 sx_padding_7; /* rr_blocks */
2029 __u32 sx_padding_8; /* rr_flags */
2030 __u32 sx_padding_9; /* rr_padding_2 */
2031 __u32 sx_padding_10; /* rr_padding_3 */
2032 __u32 sx_padding_11; /* rr_padding_4 */
2035 /* instance of mdt_reint_rec
2036 * FLR: for file resync MDS_REINT_RESYNC RPC. */
2037 struct mdt_rec_resync {
2045 __u32 rs_suppgid1_h;
2047 __u32 rs_suppgid2_h;
2048 struct lu_fid rs_fid;
2049 __u8 rs_padding0[sizeof(struct lu_fid)];
2050 struct lustre_handle rs_handle; /* rr_mtime */
2051 __s64 rs_padding1; /* rr_atime */
2052 __s64 rs_padding2; /* rr_ctime */
2053 __u64 rs_padding3; /* rr_size */
2054 __u64 rs_padding4; /* rr_blocks */
2056 __u32 rs_padding5; /* rr_mode */
2057 __u32 rs_padding6; /* rr_flags */
2058 __u32 rs_padding7; /* rr_flags_h */
2059 __u32 rs_padding8; /* rr_umask */
2060 __u32 rs_padding9; /* rr_padding_4 */
2064 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2065 * Do NOT change the size of various members, otherwise the value
2066 * will be broken in lustre_swab_mdt_rec_reint().
2068 * If you add new members in other mdt_reint_xxx structres and need to use the
2069 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2071 struct mdt_rec_reint {
2079 __u32 rr_suppgid1_h;
2081 __u32 rr_suppgid2_h;
2082 struct lu_fid rr_fid1;
2083 struct lu_fid rr_fid2;
2094 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2097 /* lmv structures */
2099 __u32 ld_tgt_count; /* how many MDS's */
2100 __u32 ld_active_tgt_count; /* how many active */
2101 __u32 ld_default_stripe_count; /* how many objects are used */
2102 __u32 ld_pattern; /* default hash pattern */
2103 __u64 ld_default_hash_size;
2104 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2105 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2106 __u32 ld_qos_maxage; /* in second */
2107 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2108 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2109 struct obd_uuid ld_uuid;
2112 /* LMV layout EA, and it will be stored both in master and slave object */
2113 struct lmv_mds_md_v1 {
2115 __u32 lmv_stripe_count;
2116 __u32 lmv_master_mdt_index; /* On master object, it is master
2117 * MDT index, on slave object, it
2118 * is stripe index of the slave obj */
2119 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2120 * which hash function to be used,
2121 * Note: only lower 16 bits is being
2122 * used for now. Higher 16 bits will
2123 * be used to mark the object status,
2124 * for example migrating or dead. */
2125 __u32 lmv_layout_version; /* Used for directory restriping */
2129 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
2130 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2133 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2134 #define LMV_MAGIC LMV_MAGIC_V1
2136 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2137 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2139 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
2140 * and the higher part will be the flag to indicate the status of object,
2141 * for example the object is being migrated. And the hash function
2142 * might be interpreted differently with different flags. */
2143 #define LMV_HASH_TYPE_MASK 0x0000ffff
2145 #define LMV_HASH_FLAG_MIGRATION 0x80000000
2147 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 11, 56, 0)
2148 /* Since lustre 2.8, this flag will not be needed, instead this DEAD
2149 * and orphan flags will be stored in LMA (see LMAI_ORPHAN)
2150 * Keep this flag just for LFSCK, because it still might meet such
2151 * flag when it checks the old FS */
2152 #define LMV_HASH_FLAG_DEAD 0x40000000
2154 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
2156 /* The striped directory has ever lost its master LMV EA, then LFSCK
2157 * re-generated it. This flag is used to indicate such case. It is an
2159 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
2162 * The FNV-1a hash algorithm is as follows:
2163 * hash = FNV_offset_basis
2164 * for each octet_of_data to be hashed
2165 * hash = hash XOR octet_of_data
2166 * hash = hash × FNV_prime
2168 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2170 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2171 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2173 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2174 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2175 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, size_t size)
2177 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2178 const unsigned char *p = buf;
2181 for (i = 0; i < size; i++) {
2183 hash *= LUSTRE_FNV_1A_64_PRIME;
2191 struct lmv_mds_md_v1 lmv_md_v1;
2192 struct lmv_user_md lmv_user_md;
2195 static inline int lmv_mds_md_size(int stripe_count, unsigned int lmm_magic)
2197 switch (lmm_magic) {
2199 struct lmv_mds_md_v1 *lmm1;
2201 return sizeof(*lmm1) + stripe_count *
2202 sizeof(lmm1->lmv_stripe_fids[0]);
2209 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2211 switch (__le32_to_cpu(lmm->lmv_magic)) {
2213 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2214 case LMV_USER_MAGIC:
2215 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2221 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2223 switch (__le32_to_cpu(lmm->lmv_magic)) {
2225 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2226 case LMV_USER_MAGIC:
2227 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2237 FLD_FIRST_OPC = FLD_QUERY
2243 SEQ_FIRST_OPC = SEQ_QUERY
2247 SEQ_ALLOC_SUPER = 0,
2259 LFSCK_NOTIFY = 1101,
2262 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2266 * LOV data structures
2269 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2270 /* The size of the buffer the lov/mdc reserves for the
2271 * array of UUIDs returned by the MDS. With the current
2272 * protocol, this will limit the max number of OSTs per LOV */
2274 #define LOV_DESC_MAGIC 0xB0CCDE5C
2275 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2276 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2278 /* LOV settings descriptor (should only contain static info) */
2280 __u32 ld_tgt_count; /* how many OBD's */
2281 __u32 ld_active_tgt_count; /* how many active */
2282 __s32 ld_default_stripe_count; /* how many objects are used */
2283 __u32 ld_pattern; /* default PATTERN_RAID0 */
2284 __u64 ld_default_stripe_size; /* in bytes */
2285 __s64 ld_default_stripe_offset; /* starting OST index */
2286 __u32 ld_padding_0; /* unused */
2287 __u32 ld_qos_maxage; /* in second */
2288 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2289 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2290 struct obd_uuid ld_uuid;
2293 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2298 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2303 LDLM_BL_CALLBACK = 104,
2304 LDLM_CP_CALLBACK = 105,
2305 LDLM_GL_CALLBACK = 106,
2306 LDLM_SET_INFO = 107,
2309 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2311 #define RES_NAME_SIZE 4
2312 struct ldlm_res_id {
2313 __u64 name[RES_NAME_SIZE];
2316 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2317 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2318 (unsigned long long)(res)->lr_name.name[1], \
2319 (unsigned long long)(res)->lr_name.name[2], \
2320 (unsigned long long)(res)->lr_name.name[3]
2323 typedef enum ldlm_mode {
2336 #define LCK_MODE_NUM 8
2338 typedef enum ldlm_type {
2346 #define LDLM_MIN_TYPE LDLM_PLAIN
2348 struct ldlm_extent {
2354 static inline bool ldlm_extent_equal(const struct ldlm_extent *ex1,
2355 const struct ldlm_extent *ex2)
2357 return ex1->start == ex2->start && ex1->end == ex2->end;
2360 struct ldlm_inodebits {
2362 __u64 try_bits; /* optional bits to try */
2365 struct ldlm_flock_wire {
2373 /* it's important that the fields of the ldlm_extent structure match
2374 * the first fields of the ldlm_flock structure because there is only
2375 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2376 * this ever changes we will need to swab the union differently based
2377 * on the resource type. */
2379 typedef union ldlm_wire_policy_data {
2380 struct ldlm_extent l_extent;
2381 struct ldlm_flock_wire l_flock;
2382 struct ldlm_inodebits l_inodebits;
2383 } ldlm_wire_policy_data_t;
2385 struct barrier_lvb {
2391 struct ldlm_gl_barrier_desc {
2397 union ldlm_gl_desc {
2398 struct ldlm_gl_lquota_desc lquota_desc;
2399 struct ldlm_gl_barrier_desc barrier_desc;
2402 enum ldlm_intent_flags {
2403 IT_OPEN = 0x00000001,
2404 IT_CREAT = 0x00000002,
2405 IT_OPEN_CREAT = 0x00000003,
2406 IT_READDIR = 0x00000004,
2407 IT_GETATTR = 0x00000008,
2408 IT_LOOKUP = 0x00000010,
2409 IT_UNLINK = 0x00000020,
2410 IT_TRUNC = 0x00000040,
2411 IT_GETXATTR = 0x00000080,
2412 IT_EXEC = 0x00000100,
2413 IT_PIN = 0x00000200,
2414 IT_LAYOUT = 0x00000400,
2415 IT_QUOTA_DQACQ = 0x00000800,
2416 IT_QUOTA_CONN = 0x00001000,
2417 IT_SETXATTR = 0x00002000,
2418 IT_GLIMPSE = 0x00004000,
2419 IT_BRW = 0x00008000,
2422 struct ldlm_intent {
2426 struct ldlm_resource_desc {
2427 enum ldlm_type lr_type;
2428 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2429 struct ldlm_res_id lr_name;
2432 struct ldlm_lock_desc {
2433 struct ldlm_resource_desc l_resource;
2434 enum ldlm_mode l_req_mode;
2435 enum ldlm_mode l_granted_mode;
2436 union ldlm_wire_policy_data l_policy_data;
2439 #define LDLM_LOCKREQ_HANDLES 2
2440 #define LDLM_ENQUEUE_CANCEL_OFF 1
2442 struct ldlm_request {
2443 __u32 lock_flags; /* LDLM_FL_*, see lustre_dlm_flags.h */
2444 __u32 lock_count; /* number of locks in lock_handle[] */
2445 struct ldlm_lock_desc lock_desc;/* lock descriptor */
2446 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2451 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2452 struct ldlm_lock_desc lock_desc;
2453 struct lustre_handle lock_handle;
2454 __u64 lock_policy_res1;
2455 __u64 lock_policy_res2;
2458 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2459 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2462 * Opcodes for mountconf (mgs and mgc)
2467 MGS_EXCEPTION, /* node died, etc. */
2468 MGS_TARGET_REG, /* whenever target starts up */
2474 #define MGS_FIRST_OPC MGS_CONNECT
2476 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 13, 53, 0)
2477 #define MGS_PARAM_MAXLEN 1024
2478 #define KEY_SET_INFO "set_info"
2480 struct mgs_send_param {
2481 char mgs_param[MGS_PARAM_MAXLEN];
2485 /* We pass this info to the MGS so it can write config logs */
2486 #define MTI_NAME_MAXLEN 64
2487 #define MTI_PARAM_MAXLEN 4096
2488 #define MTI_NIDS_MAX 32
2489 struct mgs_target_info {
2490 __u32 mti_lustre_ver;
2491 __u32 mti_stripe_index;
2492 __u32 mti_config_ver;
2493 __u32 mti_flags; /* LDD_F_* */
2494 __u32 mti_nid_count;
2495 __u32 mti_instance; /* Running instance of target */
2496 char mti_fsname[MTI_NAME_MAXLEN];
2497 char mti_svname[MTI_NAME_MAXLEN];
2498 char mti_uuid[sizeof(struct obd_uuid)];
2499 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t) */
2500 char mti_params[MTI_PARAM_MAXLEN];
2503 struct mgs_nidtbl_entry {
2504 __u64 mne_version; /* table version of this entry */
2505 __u32 mne_instance; /* target instance # */
2506 __u32 mne_index; /* target index */
2507 __u32 mne_length; /* length of this entry - by bytes */
2508 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2509 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2510 __u8 mne_nid_size; /* size of each NID, by bytes */
2511 __u8 mne_nid_count; /* # of NIDs in buffer */
2513 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2518 CONFIG_T_CONFIG = 0,
2519 CONFIG_T_SPTLRPC = 1,
2520 CONFIG_T_RECOVER = 2,
2521 CONFIG_T_PARAMS = 3,
2522 CONFIG_T_NODEMAP = 4,
2523 CONFIG_T_BARRIER = 5,
2527 struct mgs_config_body {
2528 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2529 __u64 mcb_offset; /* next index of config log to request */
2530 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2531 __u8 mcb_nm_cur_pass;
2532 __u8 mcb_bits; /* bits unit size of config log */
2533 __u32 mcb_units; /* # of units for bulk transfer */
2536 struct mgs_config_res {
2537 __u64 mcr_offset; /* index of last config log */
2539 __u64 mcr_size; /* size of the log */
2540 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2544 /* Config marker flags (in config log) */
2545 #define CM_START 0x01
2547 #define CM_SKIP 0x04
2548 #define CM_UPGRADE146 0x08
2549 #define CM_EXCLUDE 0x10
2550 #define CM_START_SKIP (CM_START | CM_SKIP)
2553 __u32 cm_step; /* aka config version */
2555 __u32 cm_vers; /* lustre release version number */
2556 __u32 cm_padding; /* 64 bit align */
2557 __s64 cm_createtime; /*when this record was first created */
2558 __s64 cm_canceltime; /*when this record is no longer valid*/
2559 char cm_tgtname[MTI_NAME_MAXLEN];
2560 char cm_comment[MTI_NAME_MAXLEN];
2564 * Opcodes for multiple servers.
2570 OBD_QC_CALLBACK, /* not used since 2.4 */
2574 #define OBD_FIRST_OPC OBD_PING
2577 * llog contexts indices.
2579 * There is compatibility problem with indexes below, they are not
2580 * continuous and must keep their numbers for compatibility needs.
2581 * See LU-5218 for details.
2584 LLOG_CONFIG_ORIG_CTXT = 0,
2585 LLOG_CONFIG_REPL_CTXT = 1,
2586 LLOG_MDS_OST_ORIG_CTXT = 2,
2587 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2588 LLOG_SIZE_ORIG_CTXT = 4,
2589 LLOG_SIZE_REPL_CTXT = 5,
2590 LLOG_TEST_ORIG_CTXT = 8,
2591 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2592 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2593 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2594 /* for multiple changelog consumers */
2595 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2596 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2597 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log */
2598 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log */
2602 /** Identifier for a single log object */
2604 struct ost_id lgl_oi;
2606 } __attribute__((packed));
2608 /** Records written to the CATALOGS list */
2609 #define CATLIST "CATALOGS"
2611 struct llog_logid lci_logid;
2615 } __attribute__((packed));
2617 /* Log data record types - there is no specific reason that these need to
2618 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2620 #define LLOG_OP_MAGIC 0x10600000
2621 #define LLOG_OP_MASK 0xfff00000
2624 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2625 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2626 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2627 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2628 REINT_UNLINK, /* obsolete after 2.5.0 */
2629 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2631 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2632 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2634 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2635 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2636 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2637 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2638 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2639 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2640 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2641 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000,
2642 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2643 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2646 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2647 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2649 /** Log record header - stored in little endian order.
2650 * Each record must start with this struct, end with a llog_rec_tail,
2651 * and be a multiple of 256 bits in size.
2653 struct llog_rec_hdr {
2660 struct llog_rec_tail {
2665 /* Where data follow just after header */
2666 #define REC_DATA(ptr) \
2667 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2669 #define REC_DATA_LEN(rec) \
2670 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2671 sizeof(struct llog_rec_tail))
2673 struct llog_logid_rec {
2674 struct llog_rec_hdr lid_hdr;
2675 struct llog_logid lid_id;
2679 struct llog_rec_tail lid_tail;
2680 } __attribute__((packed));
2682 struct llog_unlink_rec {
2683 struct llog_rec_hdr lur_hdr;
2687 struct llog_rec_tail lur_tail;
2688 } __attribute__((packed));
2690 struct llog_unlink64_rec {
2691 struct llog_rec_hdr lur_hdr;
2692 struct lu_fid lur_fid;
2693 __u32 lur_count; /* to destroy the lost precreated */
2697 struct llog_rec_tail lur_tail;
2698 } __attribute__((packed));
2700 struct llog_setattr64_rec {
2701 struct llog_rec_hdr lsr_hdr;
2702 struct ost_id lsr_oi;
2708 struct llog_rec_tail lsr_tail;
2709 } __attribute__((packed));
2711 /* Extended to support project quota */
2712 struct llog_setattr64_rec_v2 {
2713 struct llog_rec_hdr lsr_hdr;
2714 struct ost_id lsr_oi;
2721 __u32 lsr_layout_version;
2724 struct llog_rec_tail lsr_tail;
2725 } __attribute__((packed));
2727 struct llog_size_change_rec {
2728 struct llog_rec_hdr lsc_hdr;
2729 struct ll_fid lsc_fid;
2734 struct llog_rec_tail lsc_tail;
2735 } __attribute__((packed));
2737 #define CHANGELOG_MAGIC 0xca103000
2739 /** \a changelog_rec_type's that can't be masked */
2740 #define CHANGELOG_MINMASK (1 << CL_MARK)
2741 /** bits covering all \a changelog_rec_type's */
2742 #define CHANGELOG_ALLMASK 0XFFFFFFFF
2743 /** default \a changelog_rec_type mask. Allow all of them, except
2744 * CL_ATIME since it can really be time consuming, and not necessary
2745 * under normal use. */
2746 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & ~(1 << CL_ATIME))
2748 /* changelog llog name, needed by client replicators */
2749 #define CHANGELOG_CATALOG "changelog_catalog"
2751 struct changelog_setinfo {
2754 } __attribute__((packed));
2756 /** changelog record */
2757 struct llog_changelog_rec {
2758 struct llog_rec_hdr cr_hdr;
2759 struct changelog_rec cr; /**< Variable length field */
2760 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2761 } __attribute__((packed));
2763 #define CHANGELOG_USER_PREFIX "cl"
2765 struct llog_changelog_user_rec {
2766 struct llog_rec_hdr cur_hdr;
2768 /* only intended to be used in relative time comparisons to
2769 * detect idle users */
2772 struct llog_rec_tail cur_tail;
2773 } __attribute__((packed));
2775 enum agent_req_status {
2783 static inline const char *agent_req_status2name(enum agent_req_status ars)
2801 struct llog_agent_req_rec {
2802 struct llog_rec_hdr arr_hdr; /**< record header */
2803 __u32 arr_status; /**< status of the request */
2805 * agent_req_status */
2806 __u32 arr_archive_id; /**< backend archive number */
2807 __u64 arr_flags; /**< req flags */
2808 __u64 arr_compound_id; /**< compound cookie */
2809 __u64 arr_req_create; /**< req. creation time */
2810 __u64 arr_req_change; /**< req. status change time */
2811 struct hsm_action_item arr_hai; /**< req. to the agent */
2812 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2813 } __attribute__((packed));
2815 /* Old llog gen for compatibility */
2819 } __attribute__((packed));
2821 struct llog_gen_rec {
2822 struct llog_rec_hdr lgr_hdr;
2823 struct llog_gen lgr_gen;
2827 struct llog_rec_tail lgr_tail;
2830 /* flags for the logs */
2832 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2833 LLOG_F_IS_CAT = 0x2,
2834 LLOG_F_IS_PLAIN = 0x4,
2835 LLOG_F_EXT_JOBID = 0x8,
2836 LLOG_F_IS_FIXSIZE = 0x10,
2837 LLOG_F_EXT_EXTRA_FLAGS = 0x20,
2838 LLOG_F_EXT_X_UIDGID = 0x40,
2840 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
2841 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
2842 * because the catlog record is usually fixed size, but its plain
2843 * log record can be variable */
2844 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID | LLOG_F_EXT_EXTRA_FLAGS |
2845 LLOG_F_EXT_X_UIDGID,
2848 /* On-disk header structure of each log object, stored in little endian order */
2849 #define LLOG_MIN_CHUNK_SIZE 8192
2850 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
2851 * - sizeof(llh_bitmap) */
2852 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
2853 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2855 struct llog_log_hdr {
2856 struct llog_rec_hdr llh_hdr;
2857 __s64 llh_timestamp;
2859 __u32 llh_bitmap_offset;
2862 /* for a catalog the first/oldest and still in-use plain slot is just
2863 * next to it. It will serve as the upper limit after Catalog has
2866 struct obd_uuid llh_tgtuuid;
2867 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
2868 /* These fields must always be at the end of the llog_log_hdr.
2869 * Note: llh_bitmap size is variable because llog chunk size could be
2870 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
2871 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
2872 * llh_tail should only be refered by LLOG_HDR_TAIL().
2873 * But this structure is also used by client/server llog interface
2874 * (see llog_client.c), it will be kept in its original way to avoid
2875 * compatiblity issue. */
2876 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
2877 struct llog_rec_tail llh_tail;
2878 } __attribute__((packed));
2879 #undef LLOG_HEADER_SIZE
2880 #undef LLOG_BITMAP_BYTES
2882 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2883 llh->llh_bitmap_offset - \
2884 sizeof(llh->llh_tail)) * 8)
2885 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
2886 (llh)->llh_bitmap_offset)
2887 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
2888 llh->llh_hdr.lrh_len - \
2889 sizeof(llh->llh_tail)))
2891 /** log cookies are used to reference a specific log file and a record therein,
2892 and pass record offset from llog_process_thread to llog_write */
2893 struct llog_cookie {
2895 struct llog_logid lgc_lgl;
2901 } __attribute__((packed));
2903 /** llog protocol */
2904 enum llogd_rpc_ops {
2905 LLOG_ORIGIN_HANDLE_CREATE = 501,
2906 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2907 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2908 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2909 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2910 LLOG_ORIGIN_CONNECT = 506,
2911 LLOG_CATINFO = 507, /* deprecated */
2912 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2913 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2915 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2919 struct llog_logid lgd_logid;
2921 __u32 lgd_llh_flags;
2923 __u32 lgd_saved_index;
2925 __u64 lgd_cur_offset;
2926 } __attribute__((packed));
2928 struct llogd_conn_body {
2929 struct llog_gen lgdc_gen;
2930 struct llog_logid lgdc_logid;
2931 __u32 lgdc_ctxt_idx;
2932 } __attribute__((packed));
2934 /* Note: 64-bit types are 64-bit aligned in structure */
2936 __u64 o_valid; /* hot fields in this obdo */
2939 __u64 o_size; /* o_size-o_blocks == ost_lvb */
2943 __u64 o_blocks; /* brw: cli sent cached bytes */
2946 /* 32-bit fields start here: keep an even number of them via padding */
2947 __u32 o_blksize; /* optimal IO blocksize */
2948 __u32 o_mode; /* brw: cli sent cache remain */
2952 __u32 o_nlink; /* brw: checksum */
2954 __u32 o_misc; /* brw: o_dropped */
2956 __u64 o_ioepoch; /* epoch in ost writes */
2957 __u32 o_stripe_idx; /* holds stripe idx */
2959 struct lustre_handle o_handle; /* brw: lock handle to prolong
2961 /* Originally, the field is llog_cookie for destroy with unlink cookie
2962 * from MDS, it is obsolete in 2.8. Then reuse it by client to transfer
2963 * layout and PFL information in IO, setattr RPCs. Since llog_cookie is
2964 * not used on wire any longer, remove it from the obdo, then it can be
2965 * enlarged freely in the further without affect related RPCs.
2967 * sizeof(ost_layout) + sieof(__u32) == sizeof(llog_cookie). */
2968 struct ost_layout o_layout;
2969 __u32 o_layout_version;
2973 __u64 o_data_version; /* getattr: sum of iversion for
2975 * brw: grant space consumed on
2976 * the client for the write */
2978 __u32 o_padding_4; /* also fix
2979 * lustre_swab_obdo() */
2984 #define o_dirty o_blocks
2985 #define o_undirty o_mode
2986 #define o_dropped o_misc
2987 #define o_cksum o_nlink
2988 #define o_grant_used o_data_version
2990 struct lfsck_request {
3002 __u16 lr_async_windows;
3004 struct lu_fid lr_fid;
3005 struct lu_fid lr_fid2;
3013 struct lfsck_reply {
3020 LE_LASTID_REBUILDING = 1,
3021 LE_LASTID_REBUILT = 2,
3027 /* LE_FID_ACCESSED = 8, moved to lfsck_events_local */
3029 LE_CONDITIONAL_DESTROY = 10,
3030 LE_PAIRS_VERIFY = 11,
3031 LE_SET_LMV_MASTER = 15,
3032 LE_SET_LMV_SLAVE = 16,
3035 enum lfsck_event_flags {
3036 LEF_TO_OST = 0x00000001,
3037 LEF_FROM_OST = 0x00000002,
3038 LEF_SET_LMV_HASH = 0x00000004,
3039 LEF_SET_LMV_ALL = 0x00000008,
3040 LEF_RECHECK_NAME_HASH = 0x00000010,
3041 LEF_QUERY_ALL = 0x00000020,
3044 /* request structure for OST's */
3049 /* Key for FIEMAP to be used in get_info calls */
3050 struct ll_fiemap_info_key {
3052 struct obdo lfik_oa;
3053 struct fiemap lfik_fiemap;
3056 #define IDX_INFO_MAGIC 0x3D37CC37
3058 /* Index file transfer through the network. The server serializes the index into
3059 * a byte stream which is sent to the client via a bulk transfer */
3063 /* reply: see idx_info_flags below */
3066 /* request & reply: number of lu_idxpage (to be) transferred */
3070 /* request: requested attributes passed down to the iterator API */
3073 /* request & reply: index file identifier (FID) */
3074 struct lu_fid ii_fid;
3076 /* reply: version of the index file before starting to walk the index.
3077 * Please note that the version can be modified at any time during the
3081 /* request: hash to start with:
3082 * reply: hash of the first entry of the first lu_idxpage and hash
3083 * of the entry to read next if any */
3084 __u64 ii_hash_start;
3087 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3091 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3100 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3102 /* List of flags used in idx_info::ii_flags */
3103 enum idx_info_flags {
3104 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3105 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3106 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3107 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3108 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
3111 #define LIP_MAGIC 0x8A6D6B6C
3113 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3115 /* 16-byte header */
3118 __u16 lip_nr; /* number of entries in the container */
3119 __u64 lip_pad0; /* additional padding for future use */
3121 /* key/record pairs are stored in the remaining 4080 bytes.
3122 * depending upon the flags in idx_info::ii_flags, each key/record
3123 * pair might be preceded by:
3125 * - the key size (II_FL_VARKEY is set)
3126 * - the record size (II_FL_VARREC is set)
3128 * For the time being, we only support fixed-size key & record. */
3129 char lip_entries[0];
3132 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3134 /* Gather all possible type associated with a 4KB container */
3136 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3137 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3138 char lp_array[LU_PAGE_SIZE];
3141 /* security opcodes */
3144 SEC_CTX_INIT_CONT = 802,
3147 SEC_FIRST_OPC = SEC_CTX_INIT
3151 * capa related definitions
3153 #define CAPA_HMAC_MAX_LEN 64
3154 #define CAPA_HMAC_KEY_MAX_LEN 56
3156 /* NB take care when changing the sequence of elements this struct,
3157 * because the offset info is used in find_capa() */
3158 struct lustre_capa {
3159 struct lu_fid lc_fid; /** fid */
3160 __u64 lc_opc; /** operations allowed */
3161 __u64 lc_uid; /** file owner */
3162 __u64 lc_gid; /** file group */
3163 __u32 lc_flags; /** HMAC algorithm & flags */
3164 __u32 lc_keyid; /** key# used for the capability */
3165 __u32 lc_timeout; /** capa timeout value (sec) */
3166 __u32 lc_expiry; /** expiry time (sec) */
3167 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
3168 } __attribute__((packed));
3170 /** lustre_capa::lc_opc */
3172 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
3173 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
3174 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
3175 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
3176 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
3177 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
3178 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
3179 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
3180 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
3181 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
3182 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
3185 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3186 #define CAPA_OPC_MDS_ONLY \
3187 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3188 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3189 #define CAPA_OPC_OSS_ONLY \
3190 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3191 CAPA_OPC_OSS_DESTROY)
3192 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3193 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3195 /* lustre_capa::lc_hmac_alg */
3197 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
3201 #define CAPA_FL_MASK 0x00ffffff
3202 #define CAPA_HMAC_ALG_MASK 0xff000000
3204 struct lustre_capa_key {
3205 __u64 lk_seq; /**< mds# */
3206 __u32 lk_keyid; /**< key# */
3208 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
3209 } __attribute__((packed));
3211 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3212 #define LINK_EA_MAGIC 0x11EAF1DFUL
3213 struct link_ea_header {
3216 __u64 leh_len; /* total size */
3217 __u32 leh_overflow_time;
3221 /** Hardlink data is name and parent fid.
3222 * Stored in this crazy struct for maximum packing and endian-neutrality
3224 struct link_ea_entry {
3225 /** __u16 stored big-endian, unaligned */
3226 unsigned char lee_reclen[2];
3227 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3229 }__attribute__((packed));
3231 /** fid2path request/reply structure */
3232 struct getinfo_fid2path {
3233 struct lu_fid gf_fid;
3239 struct lu_fid gf_root_fid[0];
3241 } __attribute__((packed));
3243 /** path2parent request/reply structures */
3245 struct lu_fid gp_fid; /**< parent FID */
3246 __u32 gp_linkno; /**< hardlink number */
3247 __u32 gp_name_size; /**< size of the name field */
3248 char gp_name[0]; /**< zero-terminated link name */
3249 } __attribute__((packed));
3251 enum layout_intent_opc {
3252 LAYOUT_INTENT_ACCESS = 0, /** generic access */
3253 LAYOUT_INTENT_READ = 1, /** not used */
3254 LAYOUT_INTENT_WRITE = 2, /** write file, for comp layout */
3255 LAYOUT_INTENT_GLIMPSE = 3, /** not used */
3256 LAYOUT_INTENT_TRUNC = 4, /** truncate file, for comp layout */
3257 LAYOUT_INTENT_RELEASE = 5, /** reserved for HSM release */
3258 LAYOUT_INTENT_RESTORE = 6, /** reserved for HSM restore */
3261 /* enqueue layout lock with intent */
3262 struct layout_intent {
3263 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3265 struct lu_extent li_extent;
3266 } __attribute__((packed));
3269 * On the wire version of hsm_progress structure.
3271 * Contains the userspace hsm_progress and some internal fields.
3273 struct hsm_progress_kernel {
3274 /* Field taken from struct hsm_progress */
3275 struct lu_fid hpk_fid;
3277 struct hsm_extent hpk_extent;
3279 __u16 hpk_errval; /* positive val */
3281 /* Additional fields */
3282 __u64 hpk_data_version;
3284 } __attribute__((packed));
3287 * OUT_UPDATE RPC Format
3289 * During the cross-ref operation, the Master MDT, which the client send the
3290 * request to, will disassembly the operation into object updates, then OSP
3291 * will send these updates to the remote MDT to be executed.
3293 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3294 * operation and does a type of modification to an object.
3302 * update (ub_count-th)
3304 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3309 * rc [+ buffers] (1st)
3310 * rc [+ buffers] (2st)
3312 * rc [+ buffers] (nr_count-th)
3314 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3315 * be equal to ub_count.
3319 * Type of each update, if adding/deleting update, please also update
3320 * update_opcode in lustre/target/out_lib.c.
3332 OUT_INDEX_LOOKUP = 9,
3333 OUT_INDEX_INSERT = 10,
3334 OUT_INDEX_DELETE = 11,
3344 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3345 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3346 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3347 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3350 struct object_update_param {
3351 __u16 oup_len; /* length of this parameter */
3358 struct object_update {
3359 __u16 ou_type; /* enum update_type */
3360 __u16 ou_params_count; /* update parameters count */
3361 __u32 ou_result_size; /* how many bytes can return */
3362 __u32 ou_flags; /* enum update_flag */
3363 __u32 ou_padding1; /* padding 1 */
3364 __u64 ou_batchid; /* op transno on master */
3365 struct lu_fid ou_fid; /* object to be updated */
3366 struct object_update_param ou_params[0]; /* update params */
3369 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3370 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3371 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3372 /* Hold object_updates sending to the remote OUT in single RPC */
3373 struct object_update_request {
3375 __u16 ourq_count; /* number of ourq_updates[] */
3377 struct object_update ourq_updates[0];
3380 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3381 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3382 /* Header for updates request between MDTs */
3383 struct out_update_header {
3386 __u32 ouh_inline_length;
3387 __u32 ouh_reply_size;
3388 __u32 ouh_inline_data[0];
3391 struct out_update_buffer {
3396 /* the result of object update */
3397 struct object_update_result {
3404 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3405 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3406 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3407 /* Hold object_update_results being replied from the remote OUT. */
3408 struct object_update_reply {
3415 /* read update result */
3416 struct out_read_reply {
3423 /** layout swap request structure
3424 * fid1 and fid2 are in mdt_body
3426 struct mdc_swap_layouts {
3428 } __attribute__((packed));
3430 #define INLINE_RESYNC_ARRAY_SIZE 15
3431 struct close_data_resync_done {
3433 __u32 resync_ids_inline[INLINE_RESYNC_ARRAY_SIZE];
3437 struct lustre_handle cd_handle;
3438 struct lu_fid cd_fid;
3439 __u64 cd_data_version;
3441 __u64 cd_reserved[8];
3442 struct close_data_resync_done cd_resync;
3446 /* Update llog format */
3448 struct lu_fid uop_fid;
3450 __u16 uop_param_count;
3451 __u16 uop_params_off[0];
3455 struct update_op uops_op[0];
3458 struct update_params {
3459 struct object_update_param up_params[0];
3462 enum update_records_flag {
3463 UPDATE_RECORD_CONTINUE = 1 >> 0,
3466 * This is the update record format used to store the updates in
3467 * disk. All updates of the operation will be stored in ur_ops.
3468 * All of parameters for updates of the operation will be stored
3470 * To save the space of the record, parameters in ur_ops will only
3471 * remember their offset in ur_params, so to avoid storing duplicate
3472 * parameters in ur_params, which can help us save a lot space for
3473 * operation like creating striped directory.
3475 struct update_records {
3476 __u64 ur_master_transno;
3479 /* If the operation includes multiple updates, then ur_index
3480 * means the index of the update inside the whole updates. */
3482 __u32 ur_update_count;
3483 __u32 ur_param_count;
3484 struct update_ops ur_ops;
3485 /* Note ur_ops has a variable size, so comment out
3486 * the following ur_params, in case some use it directly
3487 * update_records->ur_params
3489 * struct update_params ur_params;
3493 struct llog_update_record {
3494 struct llog_rec_hdr lur_hdr;
3495 struct update_records lur_update_rec;
3496 /* Note ur_update_rec has a variable size, so comment out
3497 * the following ur_tail, in case someone use it directly
3499 * struct llog_rec_tail lur_tail;
3503 /* nodemap records, uses 32 byte record length */
3504 #define LUSTRE_NODEMAP_NAME_LENGTH 16
3505 struct nodemap_cluster_rec {
3506 char ncr_name[LUSTRE_NODEMAP_NAME_LENGTH + 1];
3510 __u32 ncr_squash_uid;
3511 __u32 ncr_squash_gid;
3514 /* lnet_nid_t is 8 bytes */
3515 struct nodemap_range_rec {
3516 lnet_nid_t nrr_start_nid;
3517 lnet_nid_t nrr_end_nid;
3522 struct nodemap_id_rec {
3530 struct nodemap_global_rec {
3541 struct nodemap_cluster_rec ncr;
3542 struct nodemap_range_rec nrr;
3543 struct nodemap_id_rec nir;
3544 struct nodemap_global_rec ngr;
3547 /* This is the lu_ladvise struct which goes out on the wire.
3548 * Corresponds to the userspace arg llapi_lu_ladvise.
3549 * value[1-4] are unspecified fields, used differently by different advices */
3551 __u16 lla_advice; /* advice type */
3552 __u16 lla_value1; /* values for different advice types */
3554 __u64 lla_start; /* first byte of extent for advice */
3555 __u64 lla_end; /* last byte of extent for advice */
3560 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3561 * arg llapi_ladvise_hdr.
3562 * value[1-3] are unspecified fields, used differently by different advices */
3563 struct ladvise_hdr {
3564 __u32 lah_magic; /* LADVISE_MAGIC */
3565 __u32 lah_count; /* number of advices */
3566 __u64 lah_flags; /* from enum ladvise_flag */
3567 __u32 lah_value1; /* unused */
3568 __u32 lah_value2; /* unused */
3569 __u64 lah_value3; /* unused */
3570 struct lu_ladvise lah_advise[0]; /* advices in this header */
3573 #if defined(__cplusplus)