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 enum lustre_msg_magic {
574 LUSTRE_MSG_MAGIC_V2 = 0x0BD00BD3,
575 LUSTRE_MSG_MAGIC_V2_SWABBED = 0xD30BD00B,
576 LUSTRE_MSG_MAGIC = LUSTRE_MSG_MAGIC_V2
579 /* flags for lm_flags */
581 MSGHDR_AT_SUPPORT = 0x1, /* adaptive timeouts, lm_cksum valid
582 * in early reply messages */
583 MSGHDR_CKSUM_INCOMPAT18 = 0x2, /* compat for 1.8, needs to be set well
584 * beyond 2.8.0 for compatibility */
587 #define lustre_msg lustre_msg_v2
588 /* we depend on this structure to be 8-byte aligned */
589 /* this type is only endian-adjusted in lustre_unpack_msg() */
590 struct lustre_msg_v2 {
591 __u32 lm_bufcount; /* number of buffers in lm_buflens[] */
592 __u32 lm_secflvr; /* 0 = no crypto, or sptlrpc security flavour */
593 __u32 lm_magic; /* RPC version magic = LUSTRE_MSG_MAGIC_V2 */
594 __u32 lm_repsize; /* size of preallocated reply buffer */
595 __u32 lm_cksum; /* CRC32 of ptlrpc_body early reply messages */
596 __u32 lm_flags; /* enum lustre_msghdr MSGHDR_* flags */
597 __u32 lm_padding_2; /* unused */
598 __u32 lm_padding_3; /* unused */
599 __u32 lm_buflens[0]; /* length of additional buffers in bytes,
600 * padded to a multiple of 8 bytes. */
602 * message buffers are packed after padded lm_buflens[] array,
603 * padded to a multiple of 8 bytes each to align contents.
607 /* ptlrpc_body packet pb_types */
608 #define PTL_RPC_MSG_REQUEST 4711 /* normal RPC request message */
609 #define PTL_RPC_MSG_ERR 4712 /* error reply if request unprocessed */
610 #define PTL_RPC_MSG_REPLY 4713 /* normal RPC reply message */
612 /* ptlrpc_body pb_version ((target_version << 16) | rpc_version) */
613 enum lustre_msg_version {
614 PTLRPC_MSG_VERSION = 0x00000003,
615 LUSTRE_VERSION_MASK = 0xffff0000,
616 LUSTRE_OBD_VERSION = 0x00010000,
617 LUSTRE_MDS_VERSION = 0x00020000,
618 LUSTRE_OST_VERSION = 0x00030000,
619 LUSTRE_DLM_VERSION = 0x00040000,
620 LUSTRE_LOG_VERSION = 0x00050000,
621 LUSTRE_MGS_VERSION = 0x00060000,
624 /* pb_flags that apply to all request messages */
625 /* #define MSG_LAST_REPLAY 0x0001 obsolete 2.0 => {REQ,LOCK}_REPLAY_DONE */
626 #define MSG_RESENT 0x0002 /* was previously sent, no reply seen */
627 #define MSG_REPLAY 0x0004 /* was processed, got reply, recovery */
628 /* #define MSG_AT_SUPPORT 0x0008 obsolete since 1.5, AT always enabled */
629 /* #define MSG_DELAY_REPLAY 0x0010 obsolete since 2.0 */
630 /* #define MSG_VERSION_REPLAY 0x0020 obsolete since 1.8.2, VBR always on */
631 #define MSG_REQ_REPLAY_DONE 0x0040 /* request replay over, locks next */
632 #define MSG_LOCK_REPLAY_DONE 0x0080 /* lock replay over, client done */
634 /* pb_op_flags for connect opcodes: MDS_CONNECT, OST_CONNECT, MGS_CONNECT */
635 #define MSG_CONNECT_RECOVERING 0x00000001 /* target is in recovery */
636 #define MSG_CONNECT_RECONNECT 0x00000002 /* tgt already has client import */
637 #define MSG_CONNECT_REPLAYABLE 0x00000004 /* target supports RPC replay */
638 /* #define MSG_CONNECT_PEER 0x00000008 obsolete since 1.2, removed in 1.5 */
639 #define MSG_CONNECT_LIBCLIENT 0x00000010 /* obsolete since 2.3, removed 2.6 */
640 #define MSG_CONNECT_INITIAL 0x00000020 /* first client connection attempt */
641 /* #define MSG_CONNECT_ASYNC 0x00000040 obsolete since 1.5 */
642 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
643 #define MSG_CONNECT_TRANSNO 0x00000100 /* client sent transno in replay */
645 /* number of previous object versions in pb_pre_versions[] */
646 #define PTLRPC_NUM_VERSIONS 4
647 /* without gss, ptlrpc_body is put at the first buffer. */
648 struct ptlrpc_body_v3 {
649 struct lustre_handle pb_handle;
650 __u32 pb_type; /* request/reply/err type: PTL_RPC_MSG_* */
651 __u32 pb_version; /* LUSTRE_*_VERSION | PTLRPC_MSG_VERSION */
652 __u32 pb_opc; /* RPC opcodes: MDS_*, OST_*, LDLM_, ... */
653 __u32 pb_status; /* negative Linux x86 error number */
654 __u64 pb_last_xid; /* highest replied XID w/o lower unreplied XID*/
655 __u16 pb_tag; /* multiple modifying RPCs virtual slot index */
658 __u64 pb_last_committed;/* rep: highest pb_transno committed to disk */
659 __u64 pb_transno; /* server-assigned transno for modifying RPCs */
660 __u32 pb_flags; /* req: MSG_* flags */
661 __u32 pb_op_flags; /* req: MSG_CONNECT_* flags */
662 __u32 pb_conn_cnt; /* connect instance of this client on server */
663 __u32 pb_timeout; /* req: max wait time; rep: service estimate */
664 __u32 pb_service_time; /* rep: server arrival to reply in seconds */
665 __u32 pb_limit; /* rep: dynamic DLM LRU lock count limit */
666 __u64 pb_slv; /* rep: dynamic DLM LRU server lock volume */
667 /* VBR: rep: previous pb_version(s) of objects modified by this RPC */
668 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
669 __u64 pb_mbits; /**< match bits for bulk request */
670 /* padding for future needs - fix lustre_swab_ptlrpc_body() also */
671 __u64 pb_padding64_0;
672 __u64 pb_padding64_1;
673 __u64 pb_padding64_2;
674 char pb_jobid[LUSTRE_JOBID_SIZE]; /* req: ASCII MPI jobid from env */
676 #define ptlrpc_body ptlrpc_body_v3
678 struct ptlrpc_body_v2 {
679 struct lustre_handle pb_handle;
684 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
685 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
688 __u64 pb_last_committed;
693 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
694 __u32 pb_service_time; /* for rep, actual service time, also used for
695 net_latency of req */
698 /* VBR: pre-versions */
699 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
700 __u64 pb_mbits; /**< unused in V2 */
701 /* padding for future needs */
702 __u64 pb_padding64_0;
703 __u64 pb_padding64_1;
704 __u64 pb_padding64_2;
707 /* message body offset for lustre_msg_v2 */
708 /* ptlrpc body offset in all request/reply messages */
709 #define MSG_PTLRPC_BODY_OFF 0
711 /* normal request/reply message record offset */
712 #define REQ_REC_OFF 1
713 #define REPLY_REC_OFF 1
715 /* ldlm request message body offset */
716 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
717 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
719 /* ldlm intent lock message body offset */
720 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
721 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
723 /* ldlm reply message body offset */
724 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
725 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
727 /** only use in req->rq_{req,rep}_swab_mask */
728 #define MSG_PTLRPC_HEADER_OFF 31
731 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
732 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
733 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
734 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
735 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
736 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
737 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
738 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
739 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
740 #define OBD_CONNECT_LARGE_ACL 0x200ULL /* more than 32 ACL entries */
741 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
742 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
743 #define OBD_CONNECT_IBITS 0x1000ULL /* not checked in 2.11+ */
744 #define OBD_CONNECT_BARRIER 0x2000ULL /* write barrier */
745 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
746 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
747 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
748 * in production. Removed in
749 * 2.9. Keep this flag to
752 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
753 * never used in production.
754 * Removed in 2.9. Keep this
755 * flag to avoid reusing.
757 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
758 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
759 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
760 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
761 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
762 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
763 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
764 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
765 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
766 #define OBD_CONNECT_REAL 0x8000000ULL /* obsolete since 2.8 */
767 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
768 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
769 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
770 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
771 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
772 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
773 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
774 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
775 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
776 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
777 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
779 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
780 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
781 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
782 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
783 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
784 * RPC error properly */
785 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
786 * finer space reservation */
787 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
788 * policy and 2.x server */
789 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
790 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
791 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
792 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
793 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
794 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* improved flock deadlock detection */
795 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/* create stripe disposition*/
796 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
798 #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */
799 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */
800 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify
802 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
803 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
804 #define OBD_CONNECT_LOCKAHEAD_OLD 0x1000000000000000ULL /* Old Cray lockahead */
806 /** bulk matchbits is sent within ptlrpc_body */
807 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL
808 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
809 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
810 /* ocd_connect_flags2 flags */
811 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* set file security context at create */
812 #define OBD_CONNECT2_LOCKAHEAD 0x2ULL /* ladvise lockahead v2 */
813 #define OBD_CONNECT2_FLR 0x20ULL /* FLR support */
816 * Please DO NOT add flag values here before first ensuring that this same
817 * flag value is not in use on some other branch. Please clear any such
818 * changes with senior engineers before starting to use a new flag. Then,
819 * submit a small patch against EVERY branch that ONLY adds the new flag,
820 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
821 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
822 * can be approved and landed easily to reserve the flag for future use. */
824 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
825 * connection. It is a temporary bug fix for Imperative Recovery interop
826 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
827 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
828 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
830 #define OCD_HAS_FLAG(ocd, flg) \
831 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
834 #ifdef HAVE_LRU_RESIZE_SUPPORT
835 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
837 #define LRU_RESIZE_CONNECT_FLAG 0
840 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
841 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
842 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
843 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
844 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
845 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
846 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
847 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
848 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
849 OBD_CONNECT_EINPROGRESS | \
850 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
851 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
852 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
853 OBD_CONNECT_FLOCK_DEAD | \
854 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
855 OBD_CONNECT_OPEN_BY_FID | \
856 OBD_CONNECT_DIR_STRIPE | OBD_CONNECT_GRANT | \
857 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_SRVLOCK | \
858 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_CKSUM | \
859 OBD_CONNECT_MULTIMODRPCS | \
860 OBD_CONNECT_SUBTREE | OBD_CONNECT_LARGE_ACL | \
861 OBD_CONNECT_GRANT_PARAM | \
862 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
864 #define MDT_CONNECT_SUPPORTED2 (OBD_CONNECT2_FILE_SECCTX | OBD_CONNECT2_FLR)
866 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
867 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
868 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
869 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
870 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
871 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
872 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
873 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
874 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
875 OBD_CONNECT_MAX_EASIZE | \
876 OBD_CONNECT_EINPROGRESS | \
877 OBD_CONNECT_JOBSTATS | \
878 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
879 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
880 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
881 OBD_CONNECT_BULK_MBITS | \
882 OBD_CONNECT_GRANT_PARAM | \
883 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
885 #define OST_CONNECT_SUPPORTED2 OBD_CONNECT2_LOCKAHEAD
887 #define ECHO_CONNECT_SUPPORTED 0
888 #define ECHO_CONNECT_SUPPORTED2 0
890 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
891 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
892 OBD_CONNECT_MNE_SWAB | OBD_CONNECT_PINGLESS |\
893 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_BARRIER)
895 #define MGS_CONNECT_SUPPORTED2 0
897 /* Features required for this version of the client to work with server */
898 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_FID | \
899 OBD_CONNECT_ATTRFID | \
902 /* This structure is used for both request and reply.
904 * If we eventually have separate connect data for different types, which we
905 * almost certainly will, then perhaps we stick a union in here. */
906 struct obd_connect_data {
907 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
908 __u32 ocd_version; /* lustre release version number */
909 __u32 ocd_grant; /* initial cache grant amount (bytes) */
910 __u32 ocd_index; /* LOV index to connect to */
911 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
912 __u64 ocd_ibits_known; /* inode bits this client understands */
913 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
914 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
915 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
916 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
917 __u64 ocd_transno; /* first transno from client to be replayed */
918 __u32 ocd_group; /* MDS group on OST */
919 __u32 ocd_cksum_types; /* supported checksum algorithms */
920 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
921 __u32 ocd_instance; /* instance # of this target */
922 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
923 /* Fields after ocd_maxbytes are only accessible by the receiver
924 * if the corresponding flag in ocd_connect_flags is set. Accessing
925 * any field after ocd_maxbytes on the receiver without a valid flag
926 * may result in out-of-bound memory access and kernel oops. */
927 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
928 __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */
929 __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */
930 __u64 ocd_connect_flags2;
931 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
932 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
933 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
934 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
935 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
936 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
937 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
938 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
939 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
940 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
941 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
942 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
943 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
946 * Please DO NOT use any fields here before first ensuring that this same
947 * field is not in use on some other branch. Please clear any such changes
948 * with senior engineers before starting to use a new field. Then, submit
949 * a small patch against EVERY branch that ONLY adds the new field along with
950 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
951 * reserve the flag for future use. */
954 * Supported checksum algorithms. Up to 32 checksum types are supported.
955 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
956 * Please update DECLARE_CKSUM_NAME/OBD_CKSUM_ALL in obd.h when adding a new
957 * algorithm and also the OBD_FL_CKSUM* flags.
959 typedef enum cksum_types {
960 OBD_CKSUM_CRC32 = 0x00000001,
961 OBD_CKSUM_ADLER = 0x00000002,
962 OBD_CKSUM_CRC32C= 0x00000004,
966 * OST requests: OBDO & OBD request records
971 OST_REPLY = 0, /* reply ? */
987 OST_QUOTACHECK = 18, /* not used since 2.4 */
989 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
991 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
993 #define OST_FIRST_OPC OST_REPLY
996 OBD_FL_INLINEDATA = 0x00000001,
997 OBD_FL_OBDMDEXISTS = 0x00000002,
998 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
999 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1000 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1001 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1002 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1003 OBD_FL_NO_PRJQUOTA = 0x00000080, /* the object's project is over
1005 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1006 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1007 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1008 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1009 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1010 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1011 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1012 OBD_FL_CKSUM_RSVD2 = 0x00008000, /* for future cksum types */
1013 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1014 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1015 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1016 * XXX: obsoleted - reserved for old
1017 * clients prior than 2.2 */
1018 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1019 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1020 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1021 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1022 /* OBD_FL_LOCAL_MASK = 0xF0000000, was local-only flags until 2.10 */
1024 /* Note that while these checksum values are currently separate bits,
1025 * in 2.x we can actually allow all values from 1-31 if we wanted. */
1026 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1027 OBD_FL_CKSUM_CRC32C,
1031 * All LOV EA magics should have the same postfix, if some new version
1032 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1033 * Lustre, even though the old version system does not recognizes such
1034 * new magic, it still can distinguish the corrupted cases by checking
1035 * the magic's postfix.
1037 #define LOV_MAGIC_MAGIC 0x0BD0
1038 #define LOV_MAGIC_MASK 0xFFFF
1040 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1041 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1042 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1043 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1044 /* reserved for specifying OSTs */
1045 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1046 #define LOV_MAGIC LOV_MAGIC_V1
1047 #define LOV_MAGIC_COMP_V1 (0x0BD60000 | LOV_MAGIC_MAGIC)
1050 * magic for fully defined striping
1051 * the idea is that we should have different magics for striping "hints"
1052 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1053 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1054 * we can't just change it w/o long way preparation, but we still need a
1055 * mechanism to allow LOD to differentiate hint versus ready striping.
1056 * so, at the moment we do a trick: MDT knows what to expect from request
1057 * depending on the case (replay uses ready striping, non-replay req uses
1058 * hints), so MDT replaces magic with appropriate one and now LOD can
1059 * easily understand what's inside -bzzz
1061 * those *_DEF magics are only used on server side internally, they
1062 * won't be put on wire or disk.
1064 #define LOV_MAGIC_DEFINED 0x10000000
1065 #define LOV_MAGIC_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V1)
1066 #define LOV_MAGIC_V3_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V3)
1067 #define LOV_MAGIC_COMP_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_COMP_V1)
1069 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1070 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1072 #define lov_ost_data lov_ost_data_v1
1073 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1074 struct ost_id l_ost_oi; /* OST object ID */
1075 __u32 l_ost_gen; /* generation of this l_ost_idx */
1076 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1079 #define lov_mds_md lov_mds_md_v1
1080 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1081 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1082 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1083 struct ost_id lmm_oi; /* LOV object ID */
1084 __u32 lmm_stripe_size; /* size of stripe in bytes */
1085 /* lmm_stripe_count used to be __u32 */
1086 __u16 lmm_stripe_count; /* num stripes in use for this object */
1087 __u16 lmm_layout_gen; /* layout generation number */
1088 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1091 #define MAX_MD_SIZE (sizeof(struct lov_mds_md) + 4 * sizeof(struct lov_ost_data))
1092 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1094 /* This is the default MDT reply size allocated, should the striping be bigger,
1095 * it will be reallocated in mdt_fix_reply.
1096 * 100 stripes is a bit less than 2.5k of data */
1097 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1098 100 * sizeof(struct lov_ost_data))
1100 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1101 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1102 #define XATTR_USER_PREFIX "user."
1103 #define XATTR_TRUSTED_PREFIX "trusted."
1104 #define XATTR_SECURITY_PREFIX "security."
1106 #define XATTR_NAME_SOM "trusted.som"
1107 #define XATTR_NAME_LOV "trusted.lov"
1108 #define XATTR_NAME_LMA "trusted.lma"
1109 #define XATTR_NAME_LMV "trusted.lmv"
1110 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1111 #define XATTR_NAME_LINK "trusted.link"
1112 #define XATTR_NAME_FID "trusted.fid"
1113 #define XATTR_NAME_VERSION "trusted.version"
1114 #define XATTR_NAME_SOM "trusted.som"
1115 #define XATTR_NAME_HSM "trusted.hsm"
1116 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1117 #define XATTR_NAME_DUMMY "trusted.dummy"
1119 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1120 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1122 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1123 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1124 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1125 struct ost_id lmm_oi; /* LOV object ID */
1126 __u32 lmm_stripe_size; /* size of stripe in bytes */
1127 /* lmm_stripe_count used to be __u32 */
1128 __u16 lmm_stripe_count; /* num stripes in use for this object */
1129 __u16 lmm_layout_gen; /* layout generation number */
1130 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1131 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1134 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1136 if (stripes == (__u16)-1)
1139 if (lmm_magic == LOV_MAGIC_V3)
1140 return sizeof(struct lov_mds_md_v3) +
1141 stripes * sizeof(struct lov_ost_data_v1);
1143 return sizeof(struct lov_mds_md_v1) +
1144 stripes * sizeof(struct lov_ost_data_v1);
1148 lov_mds_md_max_stripe_count(size_t buf_size, __u32 lmm_magic)
1150 switch (lmm_magic) {
1151 case LOV_MAGIC_V1: {
1152 struct lov_mds_md_v1 lmm;
1154 if (buf_size < sizeof(lmm))
1157 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1159 case LOV_MAGIC_V3: {
1160 struct lov_mds_md_v3 lmm;
1162 if (buf_size < sizeof(lmm))
1165 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1172 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1173 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1174 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1175 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1176 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1177 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1178 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1179 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1180 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1181 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1182 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1183 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1184 #define OBD_MD_DOM_SIZE (0X00001000ULL) /* Data-on-MDT component size */
1185 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1186 #define OBD_MD_FLGENER (0x00004000ULL) /* generation number */
1187 #define OBD_MD_LAYOUT_VERSION (0x00008000ULL) /* layout version for
1189 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1190 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1191 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1192 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1193 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1194 #define OBD_MD_FLQOS (0x00200000ULL) /* quality of service stats */
1195 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1196 #define OBD_MD_FLPRJQUOTA (0x00400000ULL) /* over quota flags sent from ost */
1197 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1198 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1199 #define OBD_MD_FLEPOCH (0x04000000ULL) /* ->ost write with ioepoch */
1200 /* ->mds if epoch opens or closes */
1201 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1202 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1203 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1204 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1205 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1207 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1208 #define OBD_MD_REINT (0x0000000200000000ULL) /* reintegrate oa */
1209 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1210 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1212 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1213 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1214 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1215 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1216 /* OBD_MD_FLRMTPERM (0x0000010000000000ULL) remote perm, obsolete */
1217 #define OBD_MD_FLMDSCAPA (0x0000020000000000ULL) /* MDS capability */
1218 #define OBD_MD_FLOSSCAPA (0x0000040000000000ULL) /* OSS capability */
1219 #define OBD_MD_FLCKSPLIT (0x0000080000000000ULL) /* Check split on server */
1220 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1221 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1222 * under lock; for xattr
1223 * requests means the
1224 * client holds the lock */
1225 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1227 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1228 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1231 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1232 #define OBD_MD_FLOSTLAYOUT (0x0080000000000000ULL) /* contain ost_layout */
1233 #define OBD_MD_FLPROJID (0x0100000000000000ULL) /* project ID */
1235 #define OBD_MD_FLALLQUOTA (OBD_MD_FLUSRQUOTA | \
1236 OBD_MD_FLGRPQUOTA | \
1239 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1240 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1241 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1242 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1243 OBD_MD_FLGENER | OBD_MD_FLRDEV | OBD_MD_FLGROUP | \
1246 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1248 /* don't forget obdo_fid which is way down at the bottom so it can
1249 * come after the definition of llog_cookie */
1253 HSS_CLEARMASK = 0x02,
1254 HSS_ARCHIVE_ID = 0x04,
1257 struct hsm_state_set {
1259 __u32 hss_archive_id;
1261 __u64 hss_clearmask;
1264 /* ost_body.data values for OST_BRW */
1266 #define OBD_BRW_READ 0x01
1267 #define OBD_BRW_WRITE 0x02
1268 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1269 #define OBD_BRW_NDELAY 0x04 /* Non-delay RPC should be issued for
1270 * this page. Non-delay RPCs have bit
1271 * rq_no_delay set. */
1272 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1273 * transfer and is not accounted in
1275 #define OBD_BRW_CHECK 0x10
1276 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1277 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1278 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1279 #define OBD_BRW_NOQUOTA 0x100
1280 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1281 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1282 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1283 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1284 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1285 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1286 * that the client is running low on
1287 * space for unstable pages; asking
1288 * it to sync quickly */
1289 #define OBD_BRW_OVER_PRJQUOTA 0x8000 /* Running out of project quota */
1291 #define OBD_BRW_OVER_ALLQUOTA (OBD_BRW_OVER_USRQUOTA | \
1292 OBD_BRW_OVER_GRPQUOTA | \
1293 OBD_BRW_OVER_PRJQUOTA)
1295 #define OBD_BRW_LOCAL1 0x80000000UL /*
1296 * osd-ldiskfs internal,
1297 * page mapped to real block
1300 #define OBD_BRW_LOCALS (OBD_BRW_LOCAL1)
1302 #define OBD_OBJECT_EOF LUSTRE_EOF
1304 #define OST_MIN_PRECREATE 32
1305 #define OST_MAX_PRECREATE 20000
1308 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1309 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1310 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1311 * high 16 bits in 2.4 and later */
1312 __u32 ioo_bufcnt; /* number of niobufs for this object */
1315 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1316 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1317 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. */
1318 #define IOOBJ_MAX_BRW_BITS 16
1319 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1320 #define ioobj_max_brw_set(ioo, num) \
1321 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1323 /* multiple of 8 bytes => can array */
1324 struct niobuf_remote {
1330 /* lock value block communicated between the filter and llite */
1332 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1333 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1334 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1335 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1336 #define OST_LVB_IS_ERR(blocks) \
1337 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1338 #define OST_LVB_SET_ERR(blocks, rc) \
1339 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1340 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1363 * lquota data structures
1366 #ifndef QUOTABLOCK_BITS
1367 # define QUOTABLOCK_BITS LUSTRE_QUOTABLOCK_BITS
1370 #ifndef QUOTABLOCK_SIZE
1371 # define QUOTABLOCK_SIZE LUSTRE_QUOTABLOCK_SIZE
1375 # define toqb lustre_stoqb
1378 /* The lquota_id structure is an union of all the possible identifier types that
1379 * can be used with quota, this includes:
1382 * - a FID which can be used for per-directory quota in the future */
1384 struct lu_fid qid_fid; /* FID for per-directory quota */
1385 __u64 qid_uid; /* user identifier */
1386 __u64 qid_gid; /* group identifier */
1387 __u64 qid_projid; /* project identifier */
1390 /* quotactl management */
1391 struct obd_quotactl {
1393 __u32 qc_type; /* see Q_* flag below */
1396 struct obd_dqinfo qc_dqinfo;
1397 struct obd_dqblk qc_dqblk;
1400 #define Q_COPY(out, in, member) (out)->member = (in)->member
1402 #define QCTL_COPY(out, in) \
1404 Q_COPY(out, in, qc_cmd); \
1405 Q_COPY(out, in, qc_type); \
1406 Q_COPY(out, in, qc_id); \
1407 Q_COPY(out, in, qc_stat); \
1408 Q_COPY(out, in, qc_dqinfo); \
1409 Q_COPY(out, in, qc_dqblk); \
1412 /* Body of quota request used for quota acquire/release RPCs between quota
1413 * master (aka QMT) and slaves (ak QSD). */
1415 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1416 * and type (data or metadata) as well as
1417 * the quota type (user or group). */
1418 union lquota_id qb_id; /* uid or gid or directory FID */
1419 __u32 qb_flags; /* see below */
1421 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1422 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1423 __u64 qb_slv_ver; /* slave index file version */
1424 struct lustre_handle qb_lockh; /* per-ID lock handle */
1425 struct lustre_handle qb_glb_lockh; /* global lock handle */
1426 __u64 qb_padding1[4];
1429 /* When the quota_body is used in the reply of quota global intent
1430 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1431 #define qb_slv_fid qb_fid
1432 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1434 #define qb_qunit qb_usage
1436 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1437 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1438 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1439 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1441 /* Quota types currently supported */
1443 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1444 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1445 LQUOTA_TYPE_PRJ = 0x02, /* maps to PRJQUOTA */
1449 /* There are 2 different resource types on which a quota limit can be enforced:
1450 * - inodes on the MDTs
1451 * - blocks on the OSTs */
1453 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1454 LQUOTA_RES_DT = 0x02,
1456 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1458 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1461 * Space accounting support
1462 * Format of an accounting record, providing disk usage information for a given
1465 struct lquota_acct_rec { /* 16 bytes */
1466 __u64 bspace; /* current space in use */
1467 __u64 ispace; /* current # inodes in use */
1471 * Global quota index support
1472 * Format of a global record, providing global quota settings for a given quota
1475 struct lquota_glb_rec { /* 32 bytes */
1476 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1477 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1478 __u64 qbr_time; /* grace time, in seconds */
1479 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1484 * Slave index support
1485 * Format of a slave record, recording how much space is granted to a given
1488 struct lquota_slv_rec { /* 8 bytes */
1489 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1490 * in #inodes or kbytes */
1493 /* Data structures associated with the quota locks */
1495 /* Glimpse descriptor used for the index & per-ID quota locks */
1496 struct ldlm_gl_lquota_desc {
1497 union lquota_id gl_id; /* quota ID subject to the glimpse */
1498 __u64 gl_flags; /* see LQUOTA_FL* below */
1499 __u64 gl_ver; /* new index version */
1500 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1501 __u64 gl_softlimit; /* new softlimit */
1505 #define gl_qunit gl_hardlimit /* current qunit value used when
1506 * glimpsing per-ID quota locks */
1508 /* quota glimpse flags */
1509 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1511 /* LVB used with quota (global and per-ID) locks */
1513 __u64 lvb_flags; /* see LQUOTA_FL* above */
1514 __u64 lvb_id_may_rel; /* space that might be released later */
1515 __u64 lvb_id_rel; /* space released by the slave for this ID */
1516 __u64 lvb_id_qunit; /* current qunit value */
1520 /* LVB used with global quota lock */
1521 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1529 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1538 MDS_GETATTR_NAME = 34,
1543 MDS_DISCONNECT = 39,
1546 MDS_PIN = 42, /* obsolete, never used in a release */
1547 MDS_UNPIN = 43, /* obsolete, never used in a release */
1549 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1551 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1554 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1556 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1558 MDS_HSM_STATE_GET = 54,
1559 MDS_HSM_STATE_SET = 55,
1560 MDS_HSM_ACTION = 56,
1561 MDS_HSM_PROGRESS = 57,
1562 MDS_HSM_REQUEST = 58,
1563 MDS_HSM_CT_REGISTER = 59,
1564 MDS_HSM_CT_UNREGISTER = 60,
1565 MDS_SWAP_LAYOUTS = 61,
1569 #define MDS_FIRST_OPC MDS_GETATTR
1572 /* opcodes for object update */
1578 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1598 /* the disposition of the intent outlines what was executed */
1599 #define DISP_IT_EXECD 0x00000001
1600 #define DISP_LOOKUP_EXECD 0x00000002
1601 #define DISP_LOOKUP_NEG 0x00000004
1602 #define DISP_LOOKUP_POS 0x00000008
1603 #define DISP_OPEN_CREATE 0x00000010
1604 #define DISP_OPEN_OPEN 0x00000020
1605 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1606 #define DISP_ENQ_OPEN_REF 0x00800000
1607 #define DISP_ENQ_CREATE_REF 0x01000000
1608 #define DISP_OPEN_LOCK 0x02000000
1609 #define DISP_OPEN_LEASE 0x04000000
1610 #define DISP_OPEN_STRIPE 0x08000000
1611 #define DISP_OPEN_DENY 0x10000000
1613 /* INODE LOCK PARTS */
1614 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
1615 * was used to protect permission (mode,
1616 * owner, group etc) before 2.4. */
1617 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1618 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1619 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1621 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
1622 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
1623 * Because for remote directories(in DNE), these locks will be granted by
1624 * different MDTs(different ldlm namespace).
1626 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
1627 * For Remote directory, the master MDT, where the remote directory is, will
1628 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
1629 * will grant LOOKUP_LOCK. */
1630 #define MDS_INODELOCK_PERM 0x000010
1631 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
1632 #define MDS_INODELOCK_DOM 0x000040 /* Data for data-on-mdt files */
1634 #define MDS_INODELOCK_MAXSHIFT 6
1635 /* This FULL lock is useful to take on unlink sort of operations */
1636 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1637 /* DOM lock shouldn't be canceled early, use this macro for ELC */
1638 #define MDS_INODELOCK_ELC (MDS_INODELOCK_FULL & ~MDS_INODELOCK_DOM)
1640 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1641 * but was moved into name[1] along with the OID to avoid consuming the
1642 * name[2,3] fields that need to be used for the quota id (also a FID). */
1644 LUSTRE_RES_ID_SEQ_OFF = 0,
1645 LUSTRE_RES_ID_VER_OID_OFF = 1,
1646 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1647 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1648 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1649 LUSTRE_RES_ID_HSH_OFF = 3
1652 #define MDS_STATUS_CONN 1
1653 #define MDS_STATUS_LOV 2
1656 /* these should be identical to their EXT4_*_FL counterparts, they are
1657 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1658 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1659 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1660 LUSTRE_APPEND_FL = 0x00000020, /* writes to file may only append */
1661 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1662 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1663 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1664 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1665 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1666 LUSTRE_DIRECTIO_FL = 0x00100000, /* Use direct i/o */
1667 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1668 LUSTRE_PROJINHERIT_FL = 0x20000000, /* Create with parents projid */
1670 /* These flags will not be identical to any EXT4_*_FL counterparts,
1671 * and only reserved for lustre purpose. Note: these flags might
1672 * be conflict with some of EXT4 flags, so
1673 * 1. these conflict flags needs to be removed when the flag is
1674 * wired by la_flags see osd_attr_get().
1675 * 2. If these flags needs to be stored into inode, they will be
1676 * stored in LMA. see LMAI_XXXX */
1677 LUSTRE_ORPHAN_FL = 0x00002000,
1678 LUSTRE_SET_SYNC_FL = 0x00040000, /* Synchronous setattr on OSTs */
1680 LUSTRE_LMA_FL_MASKS = LUSTRE_ORPHAN_FL,
1683 #ifndef FS_XFLAG_PROJINHERIT
1684 #define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
1688 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1689 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1690 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1691 * the S_* flags are kernel-internal values that change between kernel
1692 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1693 * See b=16526 for a full history. */
1694 static inline int ll_ext_to_inode_flags(int flags)
1696 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1697 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1698 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1699 #if defined(S_DIRSYNC)
1700 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1702 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0) |
1703 ((flags & LUSTRE_PROJINHERIT_FL) ? FS_XFLAG_PROJINHERIT : 0));
1706 static inline int ll_inode_to_ext_flags(int iflags)
1708 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1709 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1710 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1711 #if defined(S_DIRSYNC)
1712 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1714 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0) |
1715 ((iflags & FS_XFLAG_PROJINHERIT) ? LUSTRE_PROJINHERIT_FL : 0));
1719 /* 64 possible states */
1720 enum md_transient_state {
1721 MS_RESTORE = (1 << 0), /* restore is running */
1725 struct lu_fid mbo_fid1;
1726 struct lu_fid mbo_fid2;
1727 struct lustre_handle mbo_handle;
1729 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1733 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1735 __u64 mbo_t_state; /* transient file state defined in
1736 * enum md_transient_state
1737 * was "ino" until 2.4.0 */
1740 __u32 mbo_capability;
1744 __u32 mbo_flags; /* LUSTRE_*_FL file attributes */
1746 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1747 __u32 mbo_unused2; /* was "generation" until 2.4.0 */
1749 __u32 mbo_eadatasize;
1751 __u32 mbo_max_mdsize;
1752 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1753 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1754 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1756 __u64 mbo_dom_size; /* size of DOM component */
1757 __u64 mbo_dom_blocks; /* blocks consumed by DOM component */
1758 __u64 mbo_padding_8; /* also fix lustre_swab_mdt_body */
1759 __u64 mbo_padding_9;
1760 __u64 mbo_padding_10;
1763 struct mdt_ioepoch {
1764 struct lustre_handle mio_handle;
1765 __u64 mio_unused1; /* was ioepoch */
1766 __u32 mio_unused2; /* was flags */
1770 /* permissions for md_perm.mp_perm */
1772 CFS_SETUID_PERM = 0x01,
1773 CFS_SETGID_PERM = 0x02,
1774 CFS_SETGRP_PERM = 0x04,
1777 struct mdt_rec_setattr {
1787 __u32 sa_padding_1_h;
1788 struct lu_fid sa_fid;
1797 __u32 sa_attr_flags;
1799 __u32 sa_bias; /* some operation flags */
1806 * Attribute flags used in mdt_rec_setattr::sa_valid.
1807 * The kernel's #defines for ATTR_* should not be used over the network
1808 * since the client and MDS may run different kernels (see bug 13828)
1809 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1811 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1812 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1813 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1814 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1815 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1816 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1817 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1818 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1819 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1820 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1821 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1822 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1823 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1824 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1825 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1826 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1827 #define MDS_ATTR_PROJID 0x10000ULL /* = 65536 */
1830 MDS_CHECK_SPLIT = 1 << 0,
1831 MDS_CROSS_REF = 1 << 1,
1832 MDS_VTX_BYPASS = 1 << 2,
1833 MDS_PERM_BYPASS = 1 << 3,
1834 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1835 MDS_QUOTA_IGNORE = 1 << 5,
1836 /* Was MDS_CLOSE_CLEANUP (1 << 6), No more used */
1837 MDS_KEEP_ORPHAN = 1 << 7,
1838 MDS_RECOV_OPEN = 1 << 8,
1839 MDS_DATA_MODIFIED = 1 << 9,
1840 MDS_CREATE_VOLATILE = 1 << 10,
1841 MDS_OWNEROVERRIDE = 1 << 11,
1842 MDS_HSM_RELEASE = 1 << 12,
1843 MDS_RENAME_MIGRATE = 1 << 13,
1844 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1845 MDS_CLOSE_LAYOUT_MERGE = 1 << 15,
1846 MDS_CLOSE_RESYNC_DONE = 1 << 16,
1847 MDS_CLOSE_LAYOUT_SPLIT = 1 << 17,
1850 #define MDS_CLOSE_INTENT (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP | \
1851 MDS_CLOSE_LAYOUT_MERGE | MDS_CLOSE_LAYOUT_SPLIT | \
1852 MDS_CLOSE_RESYNC_DONE)
1854 /* instance of mdt_reint_rec */
1855 struct mdt_rec_create {
1863 __u32 cr_suppgid1_h;
1865 __u32 cr_suppgid2_h;
1866 struct lu_fid cr_fid1;
1867 struct lu_fid cr_fid2;
1868 struct lustre_handle cr_old_handle; /* handle in case of open replay */
1872 __u64 cr_padding_1; /* rr_blocks */
1875 /* use of helpers set/get_mrc_cr_flags() is needed to access
1876 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
1877 * extend cr_flags size without breaking 1.8 compat */
1878 __u32 cr_flags_l; /* for use with open, low 32 bits */
1879 __u32 cr_flags_h; /* for use with open, high 32 bits */
1880 __u32 cr_umask; /* umask for create */
1881 __u32 cr_padding_4; /* rr_padding_4 */
1884 /* instance of mdt_reint_rec */
1885 struct mdt_rec_link {
1893 __u32 lk_suppgid1_h;
1895 __u32 lk_suppgid2_h;
1896 struct lu_fid lk_fid1;
1897 struct lu_fid lk_fid2;
1899 __u64 lk_padding_1; /* rr_atime */
1900 __u64 lk_padding_2; /* rr_ctime */
1901 __u64 lk_padding_3; /* rr_size */
1902 __u64 lk_padding_4; /* rr_blocks */
1904 __u32 lk_padding_5; /* rr_mode */
1905 __u32 lk_padding_6; /* rr_flags */
1906 __u32 lk_padding_7; /* rr_padding_2 */
1907 __u32 lk_padding_8; /* rr_padding_3 */
1908 __u32 lk_padding_9; /* rr_padding_4 */
1911 /* instance of mdt_reint_rec */
1912 struct mdt_rec_unlink {
1920 __u32 ul_suppgid1_h;
1922 __u32 ul_suppgid2_h;
1923 struct lu_fid ul_fid1;
1924 struct lu_fid ul_fid2;
1926 __u64 ul_padding_2; /* rr_atime */
1927 __u64 ul_padding_3; /* rr_ctime */
1928 __u64 ul_padding_4; /* rr_size */
1929 __u64 ul_padding_5; /* rr_blocks */
1932 __u32 ul_padding_6; /* rr_flags */
1933 __u32 ul_padding_7; /* rr_padding_2 */
1934 __u32 ul_padding_8; /* rr_padding_3 */
1935 __u32 ul_padding_9; /* rr_padding_4 */
1938 /* instance of mdt_reint_rec */
1939 struct mdt_rec_rename {
1947 __u32 rn_suppgid1_h;
1949 __u32 rn_suppgid2_h;
1950 struct lu_fid rn_fid1;
1951 struct lu_fid rn_fid2;
1953 __u64 rn_padding_1; /* rr_atime */
1954 __u64 rn_padding_2; /* rr_ctime */
1955 __u64 rn_padding_3; /* rr_size */
1956 __u64 rn_padding_4; /* rr_blocks */
1957 __u32 rn_bias; /* some operation flags */
1958 __u32 rn_mode; /* cross-ref rename has mode */
1959 __u32 rn_padding_5; /* rr_flags */
1960 __u32 rn_padding_6; /* rr_padding_2 */
1961 __u32 rn_padding_7; /* rr_padding_3 */
1962 __u32 rn_padding_8; /* rr_padding_4 */
1965 /* instance of mdt_reint_rec */
1966 struct mdt_rec_setxattr {
1974 __u32 sx_suppgid1_h;
1976 __u32 sx_suppgid2_h;
1977 struct lu_fid sx_fid;
1978 __u64 sx_padding_1; /* These three are rr_fid2 */
1983 __u64 sx_padding_5; /* rr_ctime */
1984 __u64 sx_padding_6; /* rr_size */
1985 __u64 sx_padding_7; /* rr_blocks */
1988 __u32 sx_padding_8; /* rr_flags */
1989 __u32 sx_padding_9; /* rr_padding_2 */
1990 __u32 sx_padding_10; /* rr_padding_3 */
1991 __u32 sx_padding_11; /* rr_padding_4 */
1994 /* instance of mdt_reint_rec
1995 * FLR: for file resync MDS_REINT_RESYNC RPC. */
1996 struct mdt_rec_resync {
2004 __u32 rs_suppgid1_h;
2006 __u32 rs_suppgid2_h;
2007 struct lu_fid rs_fid;
2008 __u8 rs_padding0[sizeof(struct lu_fid)];
2009 struct lustre_handle rs_handle; /* rr_mtime */
2010 __s64 rs_padding1; /* rr_atime */
2011 __s64 rs_padding2; /* rr_ctime */
2012 __u64 rs_padding3; /* rr_size */
2013 __u64 rs_padding4; /* rr_blocks */
2015 __u32 rs_padding5; /* rr_mode */
2016 __u32 rs_padding6; /* rr_flags */
2017 __u32 rs_padding7; /* rr_flags_h */
2018 __u32 rs_padding8; /* rr_umask */
2019 __u32 rs_padding9; /* rr_padding_4 */
2023 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2024 * Do NOT change the size of various members, otherwise the value
2025 * will be broken in lustre_swab_mdt_rec_reint().
2027 * If you add new members in other mdt_reint_xxx structres and need to use the
2028 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2030 struct mdt_rec_reint {
2038 __u32 rr_suppgid1_h;
2040 __u32 rr_suppgid2_h;
2041 struct lu_fid rr_fid1;
2042 struct lu_fid rr_fid2;
2053 __u32 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2056 /* lmv structures */
2058 __u32 ld_tgt_count; /* how many MDS's */
2059 __u32 ld_active_tgt_count; /* how many active */
2060 __u32 ld_default_stripe_count; /* how many objects are used */
2061 __u32 ld_pattern; /* default hash pattern */
2062 __u64 ld_default_hash_size;
2063 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2064 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2065 __u32 ld_qos_maxage; /* in second */
2066 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2067 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2068 struct obd_uuid ld_uuid;
2071 /* LMV layout EA, and it will be stored both in master and slave object */
2072 struct lmv_mds_md_v1 {
2074 __u32 lmv_stripe_count;
2075 __u32 lmv_master_mdt_index; /* On master object, it is master
2076 * MDT index, on slave object, it
2077 * is stripe index of the slave obj */
2078 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2079 * which hash function to be used,
2080 * Note: only lower 16 bits is being
2081 * used for now. Higher 16 bits will
2082 * be used to mark the object status,
2083 * for example migrating or dead. */
2084 __u32 lmv_layout_version; /* Used for directory restriping */
2088 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
2089 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2092 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2093 #define LMV_MAGIC LMV_MAGIC_V1
2095 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2096 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2098 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
2099 * and the higher part will be the flag to indicate the status of object,
2100 * for example the object is being migrated. And the hash function
2101 * might be interpreted differently with different flags. */
2102 #define LMV_HASH_TYPE_MASK 0x0000ffff
2104 #define LMV_HASH_FLAG_MIGRATION 0x80000000
2106 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 11, 56, 0)
2107 /* Since lustre 2.8, this flag will not be needed, instead this DEAD
2108 * and orphan flags will be stored in LMA (see LMAI_ORPHAN)
2109 * Keep this flag just for LFSCK, because it still might meet such
2110 * flag when it checks the old FS */
2111 #define LMV_HASH_FLAG_DEAD 0x40000000
2113 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
2115 /* The striped directory has ever lost its master LMV EA, then LFSCK
2116 * re-generated it. This flag is used to indicate such case. It is an
2118 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
2121 * The FNV-1a hash algorithm is as follows:
2122 * hash = FNV_offset_basis
2123 * for each octet_of_data to be hashed
2124 * hash = hash XOR octet_of_data
2125 * hash = hash × FNV_prime
2127 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2129 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2130 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2132 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2133 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2134 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, size_t size)
2136 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2137 const unsigned char *p = buf;
2140 for (i = 0; i < size; i++) {
2142 hash *= LUSTRE_FNV_1A_64_PRIME;
2150 struct lmv_mds_md_v1 lmv_md_v1;
2151 struct lmv_user_md lmv_user_md;
2154 static inline int lmv_mds_md_size(int stripe_count, unsigned int lmm_magic)
2156 switch (lmm_magic) {
2158 struct lmv_mds_md_v1 *lmm1;
2160 return sizeof(*lmm1) + stripe_count *
2161 sizeof(lmm1->lmv_stripe_fids[0]);
2168 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2170 switch (__le32_to_cpu(lmm->lmv_magic)) {
2172 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2173 case LMV_USER_MAGIC:
2174 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2180 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2182 switch (__le32_to_cpu(lmm->lmv_magic)) {
2184 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2185 case LMV_USER_MAGIC:
2186 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2196 FLD_FIRST_OPC = FLD_QUERY
2202 SEQ_FIRST_OPC = SEQ_QUERY
2206 SEQ_ALLOC_SUPER = 0,
2218 LFSCK_NOTIFY = 1101,
2221 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2225 * LOV data structures
2228 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2229 /* The size of the buffer the lov/mdc reserves for the
2230 * array of UUIDs returned by the MDS. With the current
2231 * protocol, this will limit the max number of OSTs per LOV */
2233 #define LOV_DESC_MAGIC 0xB0CCDE5C
2234 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2235 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2237 /* LOV settings descriptor (should only contain static info) */
2239 __u32 ld_tgt_count; /* how many OBD's */
2240 __u32 ld_active_tgt_count; /* how many active */
2241 __s32 ld_default_stripe_count; /* how many objects are used */
2242 __u32 ld_pattern; /* default PATTERN_RAID0 */
2243 __u64 ld_default_stripe_size; /* in bytes */
2244 __s64 ld_default_stripe_offset; /* starting OST index */
2245 __u32 ld_padding_0; /* unused */
2246 __u32 ld_qos_maxage; /* in second */
2247 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2248 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2249 struct obd_uuid ld_uuid;
2252 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2257 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2262 LDLM_BL_CALLBACK = 104,
2263 LDLM_CP_CALLBACK = 105,
2264 LDLM_GL_CALLBACK = 106,
2265 LDLM_SET_INFO = 107,
2268 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2270 #define RES_NAME_SIZE 4
2271 struct ldlm_res_id {
2272 __u64 name[RES_NAME_SIZE];
2275 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2276 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2277 (unsigned long long)(res)->lr_name.name[1], \
2278 (unsigned long long)(res)->lr_name.name[2], \
2279 (unsigned long long)(res)->lr_name.name[3]
2282 typedef enum ldlm_mode {
2295 #define LCK_MODE_NUM 8
2297 typedef enum ldlm_type {
2305 #define LDLM_MIN_TYPE LDLM_PLAIN
2307 struct ldlm_extent {
2313 static inline bool ldlm_extent_equal(const struct ldlm_extent *ex1,
2314 const struct ldlm_extent *ex2)
2316 return ex1->start == ex2->start && ex1->end == ex2->end;
2319 struct ldlm_inodebits {
2321 __u64 try_bits; /* optional bits to try */
2324 struct ldlm_flock_wire {
2332 /* it's important that the fields of the ldlm_extent structure match
2333 * the first fields of the ldlm_flock structure because there is only
2334 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2335 * this ever changes we will need to swab the union differently based
2336 * on the resource type. */
2338 typedef union ldlm_wire_policy_data {
2339 struct ldlm_extent l_extent;
2340 struct ldlm_flock_wire l_flock;
2341 struct ldlm_inodebits l_inodebits;
2342 } ldlm_wire_policy_data_t;
2344 struct barrier_lvb {
2350 struct ldlm_gl_barrier_desc {
2356 union ldlm_gl_desc {
2357 struct ldlm_gl_lquota_desc lquota_desc;
2358 struct ldlm_gl_barrier_desc barrier_desc;
2361 enum ldlm_intent_flags {
2362 IT_OPEN = 0x00000001,
2363 IT_CREAT = 0x00000002,
2364 IT_OPEN_CREAT = 0x00000003,
2365 IT_READDIR = 0x00000004,
2366 IT_GETATTR = 0x00000008,
2367 IT_LOOKUP = 0x00000010,
2368 IT_UNLINK = 0x00000020,
2369 IT_TRUNC = 0x00000040,
2370 IT_GETXATTR = 0x00000080,
2371 IT_EXEC = 0x00000100,
2372 IT_PIN = 0x00000200,
2373 IT_LAYOUT = 0x00000400,
2374 IT_QUOTA_DQACQ = 0x00000800,
2375 IT_QUOTA_CONN = 0x00001000,
2376 IT_SETXATTR = 0x00002000,
2377 IT_GLIMPSE = 0x00004000,
2378 IT_BRW = 0x00008000,
2381 struct ldlm_intent {
2385 struct ldlm_resource_desc {
2386 enum ldlm_type lr_type;
2387 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2388 struct ldlm_res_id lr_name;
2391 struct ldlm_lock_desc {
2392 struct ldlm_resource_desc l_resource;
2393 enum ldlm_mode l_req_mode;
2394 enum ldlm_mode l_granted_mode;
2395 union ldlm_wire_policy_data l_policy_data;
2398 #define LDLM_LOCKREQ_HANDLES 2
2399 #define LDLM_ENQUEUE_CANCEL_OFF 1
2401 struct ldlm_request {
2402 __u32 lock_flags; /* LDLM_FL_*, see lustre_dlm_flags.h */
2403 __u32 lock_count; /* number of locks in lock_handle[] */
2404 struct ldlm_lock_desc lock_desc;/* lock descriptor */
2405 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2410 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2411 struct ldlm_lock_desc lock_desc;
2412 struct lustre_handle lock_handle;
2413 __u64 lock_policy_res1;
2414 __u64 lock_policy_res2;
2417 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2418 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2421 * Opcodes for mountconf (mgs and mgc)
2426 MGS_EXCEPTION, /* node died, etc. */
2427 MGS_TARGET_REG, /* whenever target starts up */
2433 #define MGS_FIRST_OPC MGS_CONNECT
2435 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 13, 53, 0)
2436 #define MGS_PARAM_MAXLEN 1024
2437 #define KEY_SET_INFO "set_info"
2439 struct mgs_send_param {
2440 char mgs_param[MGS_PARAM_MAXLEN];
2444 /* We pass this info to the MGS so it can write config logs */
2445 #define MTI_NAME_MAXLEN 64
2446 #define MTI_PARAM_MAXLEN 4096
2447 #define MTI_NIDS_MAX 32
2448 struct mgs_target_info {
2449 __u32 mti_lustre_ver;
2450 __u32 mti_stripe_index;
2451 __u32 mti_config_ver;
2452 __u32 mti_flags; /* LDD_F_* */
2453 __u32 mti_nid_count;
2454 __u32 mti_instance; /* Running instance of target */
2455 char mti_fsname[MTI_NAME_MAXLEN];
2456 char mti_svname[MTI_NAME_MAXLEN];
2457 char mti_uuid[sizeof(struct obd_uuid)];
2458 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t) */
2459 char mti_params[MTI_PARAM_MAXLEN];
2462 struct mgs_nidtbl_entry {
2463 __u64 mne_version; /* table version of this entry */
2464 __u32 mne_instance; /* target instance # */
2465 __u32 mne_index; /* target index */
2466 __u32 mne_length; /* length of this entry - by bytes */
2467 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2468 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2469 __u8 mne_nid_size; /* size of each NID, by bytes */
2470 __u8 mne_nid_count; /* # of NIDs in buffer */
2472 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2477 CONFIG_T_CONFIG = 0,
2478 CONFIG_T_SPTLRPC = 1,
2479 CONFIG_T_RECOVER = 2,
2480 CONFIG_T_PARAMS = 3,
2481 CONFIG_T_NODEMAP = 4,
2482 CONFIG_T_BARRIER = 5,
2486 struct mgs_config_body {
2487 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2488 __u64 mcb_offset; /* next index of config log to request */
2489 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2490 __u8 mcb_nm_cur_pass;
2491 __u8 mcb_bits; /* bits unit size of config log */
2492 __u32 mcb_units; /* # of units for bulk transfer */
2495 struct mgs_config_res {
2496 __u64 mcr_offset; /* index of last config log */
2498 __u64 mcr_size; /* size of the log */
2499 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2503 /* Config marker flags (in config log) */
2504 #define CM_START 0x01
2506 #define CM_SKIP 0x04
2507 #define CM_UPGRADE146 0x08
2508 #define CM_EXCLUDE 0x10
2509 #define CM_START_SKIP (CM_START | CM_SKIP)
2512 __u32 cm_step; /* aka config version */
2514 __u32 cm_vers; /* lustre release version number */
2515 __u32 cm_padding; /* 64 bit align */
2516 __s64 cm_createtime; /*when this record was first created */
2517 __s64 cm_canceltime; /*when this record is no longer valid*/
2518 char cm_tgtname[MTI_NAME_MAXLEN];
2519 char cm_comment[MTI_NAME_MAXLEN];
2523 * Opcodes for multiple servers.
2529 OBD_QC_CALLBACK, /* not used since 2.4 */
2533 #define OBD_FIRST_OPC OBD_PING
2536 * llog contexts indices.
2538 * There is compatibility problem with indexes below, they are not
2539 * continuous and must keep their numbers for compatibility needs.
2540 * See LU-5218 for details.
2543 LLOG_CONFIG_ORIG_CTXT = 0,
2544 LLOG_CONFIG_REPL_CTXT = 1,
2545 LLOG_MDS_OST_ORIG_CTXT = 2,
2546 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2547 LLOG_SIZE_ORIG_CTXT = 4,
2548 LLOG_SIZE_REPL_CTXT = 5,
2549 LLOG_TEST_ORIG_CTXT = 8,
2550 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2551 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2552 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2553 /* for multiple changelog consumers */
2554 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2555 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2556 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log */
2557 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log */
2561 /** Identifier for a single log object */
2563 struct ost_id lgl_oi;
2565 } __attribute__((packed));
2567 /** Records written to the CATALOGS list */
2568 #define CATLIST "CATALOGS"
2570 struct llog_logid lci_logid;
2574 } __attribute__((packed));
2576 /* Log data record types - there is no specific reason that these need to
2577 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2579 #define LLOG_OP_MAGIC 0x10600000
2580 #define LLOG_OP_MASK 0xfff00000
2583 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2584 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2585 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2586 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2587 REINT_UNLINK, /* obsolete after 2.5.0 */
2588 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2590 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2591 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2593 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2594 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2595 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2596 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2597 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2598 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2599 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2600 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000,
2601 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2602 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2605 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2606 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2608 /** Log record header - stored in little endian order.
2609 * Each record must start with this struct, end with a llog_rec_tail,
2610 * and be a multiple of 256 bits in size.
2612 struct llog_rec_hdr {
2619 struct llog_rec_tail {
2624 /* Where data follow just after header */
2625 #define REC_DATA(ptr) \
2626 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2628 #define REC_DATA_LEN(rec) \
2629 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2630 sizeof(struct llog_rec_tail))
2632 struct llog_logid_rec {
2633 struct llog_rec_hdr lid_hdr;
2634 struct llog_logid lid_id;
2638 struct llog_rec_tail lid_tail;
2639 } __attribute__((packed));
2641 struct llog_unlink_rec {
2642 struct llog_rec_hdr lur_hdr;
2646 struct llog_rec_tail lur_tail;
2647 } __attribute__((packed));
2649 struct llog_unlink64_rec {
2650 struct llog_rec_hdr lur_hdr;
2651 struct lu_fid lur_fid;
2652 __u32 lur_count; /* to destroy the lost precreated */
2656 struct llog_rec_tail lur_tail;
2657 } __attribute__((packed));
2659 struct llog_setattr64_rec {
2660 struct llog_rec_hdr lsr_hdr;
2661 struct ost_id lsr_oi;
2667 struct llog_rec_tail lsr_tail;
2668 } __attribute__((packed));
2670 /* Extended to support project quota */
2671 struct llog_setattr64_rec_v2 {
2672 struct llog_rec_hdr lsr_hdr;
2673 struct ost_id lsr_oi;
2680 __u32 lsr_layout_version;
2683 struct llog_rec_tail lsr_tail;
2684 } __attribute__((packed));
2686 struct llog_size_change_rec {
2687 struct llog_rec_hdr lsc_hdr;
2688 struct ll_fid lsc_fid;
2693 struct llog_rec_tail lsc_tail;
2694 } __attribute__((packed));
2696 #define CHANGELOG_MAGIC 0xca103000
2698 /** \a changelog_rec_type's that can't be masked */
2699 #define CHANGELOG_MINMASK (1 << CL_MARK)
2700 /** bits covering all \a changelog_rec_type's */
2701 #define CHANGELOG_ALLMASK 0XFFFFFFFF
2702 /** default \a changelog_rec_type mask. Allow all of them, except
2703 * CL_ATIME since it can really be time consuming, and not necessary
2705 * Remove also CL_OPEN from default list as it can be costly and only necessary
2706 * for audit purpose.
2708 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & ~(1 << CL_ATIME | 1 << CL_OPEN))
2710 /* changelog llog name, needed by client replicators */
2711 #define CHANGELOG_CATALOG "changelog_catalog"
2713 struct changelog_setinfo {
2716 } __attribute__((packed));
2718 /** changelog record */
2719 struct llog_changelog_rec {
2720 struct llog_rec_hdr cr_hdr;
2721 struct changelog_rec cr; /**< Variable length field */
2722 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2723 } __attribute__((packed));
2725 #define CHANGELOG_USER_PREFIX "cl"
2727 struct llog_changelog_user_rec {
2728 struct llog_rec_hdr cur_hdr;
2730 /* only intended to be used in relative time comparisons to
2731 * detect idle users */
2734 struct llog_rec_tail cur_tail;
2735 } __attribute__((packed));
2737 enum agent_req_status {
2745 static inline const char *agent_req_status2name(enum agent_req_status ars)
2763 struct llog_agent_req_rec {
2764 struct llog_rec_hdr arr_hdr; /**< record header */
2765 __u32 arr_status; /**< status of the request */
2767 * agent_req_status */
2768 __u32 arr_archive_id; /**< backend archive number */
2769 __u64 arr_flags; /**< req flags */
2770 __u64 arr_compound_id; /**< compound cookie */
2771 __u64 arr_req_create; /**< req. creation time */
2772 __u64 arr_req_change; /**< req. status change time */
2773 struct hsm_action_item arr_hai; /**< req. to the agent */
2774 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2775 } __attribute__((packed));
2777 /* Old llog gen for compatibility */
2781 } __attribute__((packed));
2783 struct llog_gen_rec {
2784 struct llog_rec_hdr lgr_hdr;
2785 struct llog_gen lgr_gen;
2789 struct llog_rec_tail lgr_tail;
2792 /* flags for the logs */
2794 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2795 LLOG_F_IS_CAT = 0x2,
2796 LLOG_F_IS_PLAIN = 0x4,
2797 LLOG_F_EXT_JOBID = 0x8,
2798 LLOG_F_IS_FIXSIZE = 0x10,
2799 LLOG_F_EXT_EXTRA_FLAGS = 0x20,
2800 LLOG_F_EXT_X_UIDGID = 0x40,
2801 LLOG_F_EXT_X_NID = 0x80,
2802 LLOG_F_EXT_X_OMODE = 0x100,
2804 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
2805 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
2806 * because the catlog record is usually fixed size, but its plain
2807 * log record can be variable */
2808 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID | LLOG_F_EXT_EXTRA_FLAGS |
2809 LLOG_F_EXT_X_UIDGID | LLOG_F_EXT_X_NID |
2813 /* On-disk header structure of each log object, stored in little endian order */
2814 #define LLOG_MIN_CHUNK_SIZE 8192
2815 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
2816 * - sizeof(llh_bitmap) */
2817 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
2818 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2820 struct llog_log_hdr {
2821 struct llog_rec_hdr llh_hdr;
2822 __s64 llh_timestamp;
2824 __u32 llh_bitmap_offset;
2827 /* for a catalog the first/oldest and still in-use plain slot is just
2828 * next to it. It will serve as the upper limit after Catalog has
2831 struct obd_uuid llh_tgtuuid;
2832 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
2833 /* These fields must always be at the end of the llog_log_hdr.
2834 * Note: llh_bitmap size is variable because llog chunk size could be
2835 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
2836 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
2837 * llh_tail should only be refered by LLOG_HDR_TAIL().
2838 * But this structure is also used by client/server llog interface
2839 * (see llog_client.c), it will be kept in its original way to avoid
2840 * compatiblity issue. */
2841 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
2842 struct llog_rec_tail llh_tail;
2843 } __attribute__((packed));
2844 #undef LLOG_HEADER_SIZE
2845 #undef LLOG_BITMAP_BYTES
2847 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2848 llh->llh_bitmap_offset - \
2849 sizeof(llh->llh_tail)) * 8)
2850 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
2851 (llh)->llh_bitmap_offset)
2852 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
2853 llh->llh_hdr.lrh_len - \
2854 sizeof(llh->llh_tail)))
2856 /** log cookies are used to reference a specific log file and a record therein,
2857 and pass record offset from llog_process_thread to llog_write */
2858 struct llog_cookie {
2860 struct llog_logid lgc_lgl;
2866 } __attribute__((packed));
2868 /** llog protocol */
2869 enum llogd_rpc_ops {
2870 LLOG_ORIGIN_HANDLE_CREATE = 501,
2871 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2872 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2873 LLOG_ORIGIN_HANDLE_WRITE_REC = 504,
2874 LLOG_ORIGIN_HANDLE_CLOSE = 505,
2875 LLOG_ORIGIN_CONNECT = 506,
2876 LLOG_CATINFO = 507, /* deprecated */
2877 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2878 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* for destroy llog object*/
2880 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2884 struct llog_logid lgd_logid;
2886 __u32 lgd_llh_flags;
2888 __u32 lgd_saved_index;
2890 __u64 lgd_cur_offset;
2891 } __attribute__((packed));
2893 struct llogd_conn_body {
2894 struct llog_gen lgdc_gen;
2895 struct llog_logid lgdc_logid;
2896 __u32 lgdc_ctxt_idx;
2897 } __attribute__((packed));
2899 /* Note: 64-bit types are 64-bit aligned in structure */
2901 __u64 o_valid; /* hot fields in this obdo */
2904 __u64 o_size; /* o_size-o_blocks == ost_lvb */
2908 __u64 o_blocks; /* brw: cli sent cached bytes */
2911 /* 32-bit fields start here: keep an even number of them via padding */
2912 __u32 o_blksize; /* optimal IO blocksize */
2913 __u32 o_mode; /* brw: cli sent cache remain */
2917 __u32 o_nlink; /* brw: checksum */
2919 __u32 o_misc; /* brw: o_dropped */
2921 __u64 o_ioepoch; /* epoch in ost writes */
2922 __u32 o_stripe_idx; /* holds stripe idx */
2924 struct lustre_handle o_handle; /* brw: lock handle to prolong
2926 /* Originally, the field is llog_cookie for destroy with unlink cookie
2927 * from MDS, it is obsolete in 2.8. Then reuse it by client to transfer
2928 * layout and PFL information in IO, setattr RPCs. Since llog_cookie is
2929 * not used on wire any longer, remove it from the obdo, then it can be
2930 * enlarged freely in the further without affect related RPCs.
2932 * sizeof(ost_layout) + sieof(__u32) == sizeof(llog_cookie). */
2933 struct ost_layout o_layout;
2934 __u32 o_layout_version;
2938 __u64 o_data_version; /* getattr: sum of iversion for
2940 * brw: grant space consumed on
2941 * the client for the write */
2943 __u32 o_padding_4; /* also fix
2944 * lustre_swab_obdo() */
2949 #define o_dirty o_blocks
2950 #define o_undirty o_mode
2951 #define o_dropped o_misc
2952 #define o_cksum o_nlink
2953 #define o_grant_used o_data_version
2955 struct lfsck_request {
2967 __u16 lr_async_windows;
2969 struct lu_fid lr_fid;
2970 struct lu_fid lr_fid2;
2978 struct lfsck_reply {
2985 LE_LASTID_REBUILDING = 1,
2986 LE_LASTID_REBUILT = 2,
2992 /* LE_FID_ACCESSED = 8, moved to lfsck_events_local */
2994 LE_CONDITIONAL_DESTROY = 10,
2995 LE_PAIRS_VERIFY = 11,
2996 LE_SET_LMV_MASTER = 15,
2997 LE_SET_LMV_SLAVE = 16,
3000 enum lfsck_event_flags {
3001 LEF_TO_OST = 0x00000001,
3002 LEF_FROM_OST = 0x00000002,
3003 LEF_SET_LMV_HASH = 0x00000004,
3004 LEF_SET_LMV_ALL = 0x00000008,
3005 LEF_RECHECK_NAME_HASH = 0x00000010,
3006 LEF_QUERY_ALL = 0x00000020,
3009 /* request structure for OST's */
3014 /* Key for FIEMAP to be used in get_info calls */
3015 struct ll_fiemap_info_key {
3017 struct obdo lfik_oa;
3018 struct fiemap lfik_fiemap;
3021 #define IDX_INFO_MAGIC 0x3D37CC37
3023 /* Index file transfer through the network. The server serializes the index into
3024 * a byte stream which is sent to the client via a bulk transfer */
3028 /* reply: see idx_info_flags below */
3031 /* request & reply: number of lu_idxpage (to be) transferred */
3035 /* request: requested attributes passed down to the iterator API */
3038 /* request & reply: index file identifier (FID) */
3039 struct lu_fid ii_fid;
3041 /* reply: version of the index file before starting to walk the index.
3042 * Please note that the version can be modified at any time during the
3046 /* request: hash to start with:
3047 * reply: hash of the first entry of the first lu_idxpage and hash
3048 * of the entry to read next if any */
3049 __u64 ii_hash_start;
3052 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3056 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3065 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3067 /* List of flags used in idx_info::ii_flags */
3068 enum idx_info_flags {
3069 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3070 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3071 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3072 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3073 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
3076 #define LIP_MAGIC 0x8A6D6B6C
3078 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3080 /* 16-byte header */
3083 __u16 lip_nr; /* number of entries in the container */
3084 __u64 lip_pad0; /* additional padding for future use */
3086 /* key/record pairs are stored in the remaining 4080 bytes.
3087 * depending upon the flags in idx_info::ii_flags, each key/record
3088 * pair might be preceded by:
3090 * - the key size (II_FL_VARKEY is set)
3091 * - the record size (II_FL_VARREC is set)
3093 * For the time being, we only support fixed-size key & record. */
3094 char lip_entries[0];
3097 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3099 /* Gather all possible type associated with a 4KB container */
3101 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3102 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3103 char lp_array[LU_PAGE_SIZE];
3106 /* security opcodes */
3109 SEC_CTX_INIT_CONT = 802,
3112 SEC_FIRST_OPC = SEC_CTX_INIT
3116 * capa related definitions
3118 #define CAPA_HMAC_MAX_LEN 64
3119 #define CAPA_HMAC_KEY_MAX_LEN 56
3121 /* NB take care when changing the sequence of elements this struct,
3122 * because the offset info is used in find_capa() */
3123 struct lustre_capa {
3124 struct lu_fid lc_fid; /** fid */
3125 __u64 lc_opc; /** operations allowed */
3126 __u64 lc_uid; /** file owner */
3127 __u64 lc_gid; /** file group */
3128 __u32 lc_flags; /** HMAC algorithm & flags */
3129 __u32 lc_keyid; /** key# used for the capability */
3130 __u32 lc_timeout; /** capa timeout value (sec) */
3131 __u32 lc_expiry; /** expiry time (sec) */
3132 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
3133 } __attribute__((packed));
3135 /** lustre_capa::lc_opc */
3137 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
3138 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
3139 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
3140 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
3141 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
3142 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
3143 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
3144 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
3145 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
3146 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
3147 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
3150 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3151 #define CAPA_OPC_MDS_ONLY \
3152 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3153 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3154 #define CAPA_OPC_OSS_ONLY \
3155 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3156 CAPA_OPC_OSS_DESTROY)
3157 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3158 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3160 /* lustre_capa::lc_hmac_alg */
3162 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
3166 #define CAPA_FL_MASK 0x00ffffff
3167 #define CAPA_HMAC_ALG_MASK 0xff000000
3169 struct lustre_capa_key {
3170 __u64 lk_seq; /**< mds# */
3171 __u32 lk_keyid; /**< key# */
3173 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
3174 } __attribute__((packed));
3176 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3177 #define LINK_EA_MAGIC 0x11EAF1DFUL
3178 struct link_ea_header {
3181 __u64 leh_len; /* total size */
3182 __u32 leh_overflow_time;
3186 /** Hardlink data is name and parent fid.
3187 * Stored in this crazy struct for maximum packing and endian-neutrality
3189 struct link_ea_entry {
3190 /** __u16 stored big-endian, unaligned */
3191 unsigned char lee_reclen[2];
3192 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3194 }__attribute__((packed));
3196 /** fid2path request/reply structure */
3197 struct getinfo_fid2path {
3198 struct lu_fid gf_fid;
3204 struct lu_fid gf_root_fid[0];
3206 } __attribute__((packed));
3208 /** path2parent request/reply structures */
3210 struct lu_fid gp_fid; /**< parent FID */
3211 __u32 gp_linkno; /**< hardlink number */
3212 __u32 gp_name_size; /**< size of the name field */
3213 char gp_name[0]; /**< zero-terminated link name */
3214 } __attribute__((packed));
3216 enum layout_intent_opc {
3217 LAYOUT_INTENT_ACCESS = 0, /** generic access */
3218 LAYOUT_INTENT_READ = 1, /** not used */
3219 LAYOUT_INTENT_WRITE = 2, /** write file, for comp layout */
3220 LAYOUT_INTENT_GLIMPSE = 3, /** not used */
3221 LAYOUT_INTENT_TRUNC = 4, /** truncate file, for comp layout */
3222 LAYOUT_INTENT_RELEASE = 5, /** reserved for HSM release */
3223 LAYOUT_INTENT_RESTORE = 6, /** reserved for HSM restore */
3226 /* enqueue layout lock with intent */
3227 struct layout_intent {
3228 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3230 struct lu_extent li_extent;
3231 } __attribute__((packed));
3234 * On the wire version of hsm_progress structure.
3236 * Contains the userspace hsm_progress and some internal fields.
3238 struct hsm_progress_kernel {
3239 /* Field taken from struct hsm_progress */
3240 struct lu_fid hpk_fid;
3242 struct hsm_extent hpk_extent;
3244 __u16 hpk_errval; /* positive val */
3246 /* Additional fields */
3247 __u64 hpk_data_version;
3249 } __attribute__((packed));
3252 * OUT_UPDATE RPC Format
3254 * During the cross-ref operation, the Master MDT, which the client send the
3255 * request to, will disassembly the operation into object updates, then OSP
3256 * will send these updates to the remote MDT to be executed.
3258 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3259 * operation and does a type of modification to an object.
3267 * update (ub_count-th)
3269 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3274 * rc [+ buffers] (1st)
3275 * rc [+ buffers] (2st)
3277 * rc [+ buffers] (nr_count-th)
3279 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3280 * be equal to ub_count.
3284 * Type of each update, if adding/deleting update, please also update
3285 * update_opcode in lustre/target/out_lib.c.
3297 OUT_INDEX_LOOKUP = 9,
3298 OUT_INDEX_INSERT = 10,
3299 OUT_INDEX_DELETE = 11,
3309 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3310 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3311 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3312 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3315 struct object_update_param {
3316 __u16 oup_len; /* length of this parameter */
3323 struct object_update {
3324 __u16 ou_type; /* enum update_type */
3325 __u16 ou_params_count; /* update parameters count */
3326 __u32 ou_result_size; /* how many bytes can return */
3327 __u32 ou_flags; /* enum update_flag */
3328 __u32 ou_padding1; /* padding 1 */
3329 __u64 ou_batchid; /* op transno on master */
3330 struct lu_fid ou_fid; /* object to be updated */
3331 struct object_update_param ou_params[0]; /* update params */
3334 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3335 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3336 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3337 /* Hold object_updates sending to the remote OUT in single RPC */
3338 struct object_update_request {
3340 __u16 ourq_count; /* number of ourq_updates[] */
3342 struct object_update ourq_updates[0];
3345 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3346 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3347 /* Header for updates request between MDTs */
3348 struct out_update_header {
3351 __u32 ouh_inline_length;
3352 __u32 ouh_reply_size;
3353 __u32 ouh_inline_data[0];
3356 struct out_update_buffer {
3361 /* the result of object update */
3362 struct object_update_result {
3369 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3370 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3371 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3372 /* Hold object_update_results being replied from the remote OUT. */
3373 struct object_update_reply {
3380 /* read update result */
3381 struct out_read_reply {
3388 /** layout swap request structure
3389 * fid1 and fid2 are in mdt_body
3391 struct mdc_swap_layouts {
3393 } __attribute__((packed));
3395 #define INLINE_RESYNC_ARRAY_SIZE 15
3396 struct close_data_resync_done {
3398 __u32 resync_ids_inline[INLINE_RESYNC_ARRAY_SIZE];
3402 struct lustre_handle cd_handle;
3403 struct lu_fid cd_fid;
3404 __u64 cd_data_version;
3406 __u64 cd_reserved[8];
3407 struct close_data_resync_done cd_resync;
3413 /* Update llog format */
3415 struct lu_fid uop_fid;
3417 __u16 uop_param_count;
3418 __u16 uop_params_off[0];
3422 struct update_op uops_op[0];
3425 struct update_params {
3426 struct object_update_param up_params[0];
3429 enum update_records_flag {
3430 UPDATE_RECORD_CONTINUE = 1 >> 0,
3433 * This is the update record format used to store the updates in
3434 * disk. All updates of the operation will be stored in ur_ops.
3435 * All of parameters for updates of the operation will be stored
3437 * To save the space of the record, parameters in ur_ops will only
3438 * remember their offset in ur_params, so to avoid storing duplicate
3439 * parameters in ur_params, which can help us save a lot space for
3440 * operation like creating striped directory.
3442 struct update_records {
3443 __u64 ur_master_transno;
3446 /* If the operation includes multiple updates, then ur_index
3447 * means the index of the update inside the whole updates. */
3449 __u32 ur_update_count;
3450 __u32 ur_param_count;
3451 struct update_ops ur_ops;
3452 /* Note ur_ops has a variable size, so comment out
3453 * the following ur_params, in case some use it directly
3454 * update_records->ur_params
3456 * struct update_params ur_params;
3460 struct llog_update_record {
3461 struct llog_rec_hdr lur_hdr;
3462 struct update_records lur_update_rec;
3463 /* Note ur_update_rec has a variable size, so comment out
3464 * the following ur_tail, in case someone use it directly
3466 * struct llog_rec_tail lur_tail;
3470 /* nodemap records, uses 32 byte record length */
3471 #define LUSTRE_NODEMAP_NAME_LENGTH 16
3472 struct nodemap_cluster_rec {
3473 char ncr_name[LUSTRE_NODEMAP_NAME_LENGTH + 1];
3477 __u32 ncr_squash_uid;
3478 __u32 ncr_squash_gid;
3481 /* lnet_nid_t is 8 bytes */
3482 struct nodemap_range_rec {
3483 lnet_nid_t nrr_start_nid;
3484 lnet_nid_t nrr_end_nid;
3489 struct nodemap_id_rec {
3497 struct nodemap_global_rec {
3508 struct nodemap_cluster_rec ncr;
3509 struct nodemap_range_rec nrr;
3510 struct nodemap_id_rec nir;
3511 struct nodemap_global_rec ngr;
3514 /* This is the lu_ladvise struct which goes out on the wire.
3515 * Corresponds to the userspace arg llapi_lu_ladvise.
3516 * value[1-4] are unspecified fields, used differently by different advices */
3518 __u16 lla_advice; /* advice type */
3519 __u16 lla_value1; /* values for different advice types */
3521 __u64 lla_start; /* first byte of extent for advice */
3522 __u64 lla_end; /* last byte of extent for advice */
3527 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3528 * arg llapi_ladvise_hdr.
3529 * value[1-3] are unspecified fields, used differently by different advices */
3530 struct ladvise_hdr {
3531 __u32 lah_magic; /* LADVISE_MAGIC */
3532 __u32 lah_count; /* number of advices */
3533 __u64 lah_flags; /* from enum ladvise_flag */
3534 __u32 lah_value1; /* unused */
3535 __u32 lah_value2; /* unused */
3536 __u64 lah_value3; /* unused */
3537 struct lu_ladvise lah_advise[0]; /* advices in this header */
3540 #if defined(__cplusplus)