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 can use any nonnegative integer to represent archive-Ids during
177 * register with MDT thru kuc.
178 * archive num = 0 => all
179 * archive num from 1 to MAX_U32
181 #define LL_HSM_ORIGIN_MAX_ARCHIVE (sizeof(__u32) * 8)
182 /* the max count of archive ids that one agent can support */
183 #define LL_HSM_MAX_ARCHIVES_PER_AGENT 1024
186 * HSM on-disk attributes stored in a separate xattr.
189 /** Bitfield for supported data in this structure. For future use. */
192 /** HSM flags, see hsm_flags enum below */
194 /** backend archive id associated with the file */
196 /** version associated with the last archiving, if any */
199 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
205 /** LASTID file has zero OID */
206 LUSTRE_FID_LASTID_OID = 0UL,
207 /** initial fid id value */
208 LUSTRE_FID_INIT_OID = 1UL
212 * Different FID Format
213 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
216 * File IDentifier generated by client from range allocated by the seq service.
217 * First 0x400 sequences [2^33, 2^33 + 0x400] are reserved for system use. Note
218 * that on ldiskfs MDTs that IGIF FIDs can use inode numbers starting at 12,
219 * but this is in the IGIF SEQ rangeand does not conflict with assigned FIDs.
222 * Inode and Generation In FID, a surrogate FID used to globally identify an
223 * existing object on OLD formatted MDT file system. This would only be used on
224 * MDT0 in a DNE filesystem, because there are not expected to be any OLD
225 * formatted DNE filesystems. Belongs to a sequence in [12, 2^32 - 1] range,
226 * where sequence number is inode number, and inode generation is used as OID.
227 * NOTE: This assumes no more than 2^32-1 inodes exist in the MDT filesystem,
228 * which is the maximum possible for an ldiskfs backend. NOTE: This assumes
229 * that the reserved ext3/ext4/ldiskfs inode numbers [0-11] are never visible
230 * to clients, which has always been true.
233 * Object ID in FID, a surrogate FID used to globally identify an existing
234 * object on OLD formatted OST file system. Belongs to a sequence in
235 * [2^32, 2^33 - 1]. Sequence number is calculated as:
236 * 1 << 32 | (ost_index << 16) | ((objid >> 32) & 0xffff)
237 * that is, SEQ consists of 16-bit OST index, and higher 16 bits of object ID.
238 * The generation of unique SEQ values per OST allows the IDIF FIDs to be
239 * identified in the FLD correctly. The OID field is calculated as:
241 * that is, it consists of lower 32 bits of object ID. NOTE This assumes that
242 * no more than 2^48-1 objects have ever been created on an OST, and that no
243 * more than 65535 OSTs are in use. Both are very reasonable assumptions (can
244 * uniquely map all objects on an OST that created 1M objects per second for 9
245 * years, or combinations thereof).
248 * Surrogate FID used to identify an existing object on OLD formatted OST
249 * filesystem. Belongs to the reserved sequence 0, and is used internally prior
250 * to the introduction of FID-on-OST, at which point IDIF will be used to
251 * identify objects as residing on a specific OST.
254 * For Lustre Log objects the object sequence 1 is used. This is compatible with
255 * both OLD and NEW.1 namespaces, as this SEQ number is in the ext3/ldiskfs
256 * reserved inode range and does not conflict with IGIF sequence numbers.
259 * For testing OST IO performance the object sequence 2 is used. This is
260 * compatible with both OLD and NEW.1 namespaces, as this SEQ number is in the
261 * ext3/ldiskfs reserved inode range and does not conflict with IGIF sequence
264 * OST_MDT1 .. OST_MAX:
265 * For testing with multiple MDTs the object sequence 3 through 9 is used,
266 * allowing direct mapping of MDTs 1 through 7 respectively, for a total of 8
267 * MDTs including OST_MDT0. This matches the legacy CMD project "group"
268 * mappings. However, this SEQ range is only for testing prior to any production
269 * DNE release, as the objects in this range conflict across all OSTs, as the
270 * OST index is not part of the FID.
273 * For compatibility with existing OLD OST network protocol structures, the FID
274 * must map onto the o_id and o_gr in a manner that ensures existing objects are
275 * identified consistently for IO, as well as onto the lock namespace to ensure
276 * both IDIFs map onto the same objects for IO as well as resources in the DLM.
279 * resource[] = {o_id, o_seq, 0, 0}; // o_seq == 0 for production releases
281 * DLM NEW.1 FID (this is the same for both the MDT and OST):
282 * resource[] = {SEQ, OID, VER, HASH};
284 * Note that for mapping IDIF values to DLM resource names the o_id may be
285 * larger than the 2^33 reserved sequence numbers for IDIF, so it is possible
286 * for the o_id numbers to overlap FID SEQ numbers in the resource. However, in
287 * all production releases the OLD o_seq field is always zero, and all valid FID
288 * OID values are non-zero, so the lock resources will not collide.
290 * For objects within the IDIF range, group extraction (non-CMD) will be:
291 * o_id = (fid->f_seq & 0x7fff) << 16 | fid->f_oid;
292 * o_seq = 0; // formerly group number
296 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
297 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
298 * used for other purposes and not risk collisions with existing inodes.
301 FID_SEQ_OST_MDT0 = 0,
302 FID_SEQ_LLOG = 1, /* unnamed llogs */
304 FID_SEQ_UNUSED_START = 3,
305 FID_SEQ_UNUSED_END = 9,
306 FID_SEQ_LLOG_NAME = 10, /* named llogs */
309 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
310 FID_SEQ_IDIF = 0x100000000ULL,
311 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
312 /* Normal FID sequence starts from this value, i.e. 1<<33 */
313 FID_SEQ_START = 0x200000000ULL,
314 /* sequence for local pre-defined FIDs listed in local_oid */
315 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
316 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
317 /* sequence is used for local named objects FIDs generated
318 * by local_object_storage library */
319 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
320 /* Because current FLD will only cache the fid sequence, instead
321 * of oid on the client side, if the FID needs to be exposed to
322 * clients sides, it needs to make sure all of fids under one
323 * sequence will be located in one MDT. */
324 FID_SEQ_SPECIAL = 0x200000004ULL,
325 FID_SEQ_QUOTA = 0x200000005ULL,
326 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
327 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
328 FID_SEQ_LAYOUT_RBTREE = 0x200000008ULL,
329 /* sequence is used for update logs of cross-MDT operation */
330 FID_SEQ_UPDATE_LOG = 0x200000009ULL,
331 /* Sequence is used for the directory under which update logs
333 FID_SEQ_UPDATE_LOG_DIR = 0x20000000aULL,
334 FID_SEQ_NORMAL = 0x200000400ULL,
335 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
338 #define OBIF_OID_MAX_BITS 32
339 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
340 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
341 #define IDIF_OID_MAX_BITS 48
342 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
343 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
345 /** OID for FID_SEQ_SPECIAL */
347 /* Big Filesystem Lock to serialize rename operations */
348 FID_OID_SPECIAL_BFL = 1UL,
351 /** OID for FID_SEQ_DOT_LUSTRE */
352 enum dot_lustre_oid {
353 FID_OID_DOT_LUSTRE = 1UL,
354 FID_OID_DOT_LUSTRE_OBF = 2UL,
355 FID_OID_DOT_LUSTRE_LPF = 3UL,
358 /** OID for FID_SEQ_ROOT */
361 FID_OID_ECHO_ROOT = 2UL,
364 struct lu_orphan_rec {
365 /* The MDT-object's FID referenced by the orphan OST-object */
366 struct lu_fid lor_fid;
371 struct lu_orphan_ent {
372 /* The orphan OST-object's FID */
373 struct lu_fid loe_key;
374 struct lu_orphan_rec loe_rec;
377 struct lu_orphan_rec_v2 {
378 struct lu_orphan_rec lor_rec;
379 struct ost_layout lor_layout;
383 struct lu_orphan_ent_v2 {
384 /* The orphan OST-object's FID */
385 struct lu_fid loe_key;
386 struct lu_orphan_rec_v2 loe_rec;
389 struct lu_orphan_rec_v3 {
390 struct lu_orphan_rec lor_rec;
391 struct ost_layout lor_layout;
392 /* The OST-object declared layout version in PFID EA.*/
393 __u32 lor_layout_version;
394 /* The OST-object declared layout range (of version) in PFID EA.*/
400 struct lu_orphan_ent_v3 {
401 /* The orphan OST-object's FID */
402 struct lu_fid loe_key;
403 struct lu_orphan_rec_v3 loe_rec;
408 /** \defgroup lu_dir lu_dir
412 * Enumeration of possible directory entry attributes.
414 * Attributes follow directory entry header in the order they appear in this
417 enum lu_dirent_attrs {
420 LUDA_64BITHASH = 0x0004,
422 /* The following attrs are used for MDT internal only,
423 * not visible to client */
425 /* Something in the record is unknown, to be verified in further. */
426 LUDA_UNKNOWN = 0x0400,
427 /* Ignore this record, go to next directly. */
428 LUDA_IGNORE = 0x0800,
429 /* The system is upgraded, has beed or to be repaired (dryrun). */
430 LUDA_UPGRADE = 0x1000,
431 /* The dirent has been repaired, or to be repaired (dryrun). */
432 LUDA_REPAIR = 0x2000,
433 /* Only check but not repair the dirent inconsistency */
434 LUDA_VERIFY_DRYRUN = 0x4000,
435 /* Verify the dirent consistency */
436 LUDA_VERIFY = 0x8000,
439 #define LU_DIRENT_ATTRS_MASK 0xff00
442 * Layout of readdir pages, as transmitted on wire.
445 /** valid if LUDA_FID is set. */
446 struct lu_fid lde_fid;
447 /** a unique entry identifier: a hash or an offset. */
449 /** total record length, including all attributes. */
453 /** optional variable size attributes following this entry.
454 * taken from enum lu_dirent_attrs.
457 /** name is followed by the attributes indicated in ->ldp_attrs, in
458 * their natural order. After the last attribute, padding bytes are
459 * added to make ->lde_reclen a multiple of 8.
465 * Definitions of optional directory entry attributes formats.
467 * Individual attributes do not have their length encoded in a generic way. It
468 * is assumed that consumer of an attribute knows its format. This means that
469 * it is impossible to skip over an unknown attribute, except by skipping over all
470 * remaining attributes (by using ->lde_reclen), which is not too
471 * constraining, because new server versions will append new attributes at
472 * the end of an entry.
476 * Fid directory attribute: a fid of an object referenced by the entry. This
477 * will be almost always requested by the client and supplied by the server.
479 * Aligned to 8 bytes.
481 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
486 * Aligned to 2 bytes.
493 __u64 ldp_hash_start;
497 struct lu_dirent ldp_entries[0];
500 enum lu_dirpage_flags {
502 * dirpage contains no entry.
506 * last entry's lde_hash equals ldp_hash_end.
511 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
513 if (__le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
516 return dp->ldp_entries;
519 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
521 struct lu_dirent *next;
523 if (__le16_to_cpu(ent->lde_reclen) != 0)
524 next = ((void *)ent) + __le16_to_cpu(ent->lde_reclen);
531 static inline size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
535 if (attr & LUDA_TYPE) {
536 const size_t align = sizeof(struct luda_type) - 1;
538 size = (sizeof(struct lu_dirent) + namelen + 1 + align) &
540 size += sizeof(struct luda_type);
542 size = sizeof(struct lu_dirent) + namelen + 1;
545 return (size + 7) & ~7;
548 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
551 * MDS_READPAGE page size
553 * This is the directory page size packed in MDS_READPAGE RPC.
554 * It's different than PAGE_SIZE because the client needs to
555 * access the struct lu_dirpage header packed at the beginning of
556 * the "page" and without this there isn't any way to know find the
557 * lu_dirpage header is if client and server PAGE_SIZE differ.
559 #define LU_PAGE_SHIFT 12
560 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
561 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
563 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
567 struct lustre_handle {
570 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
572 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
574 return lh->cookie != 0;
577 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
578 const struct lustre_handle *lh2)
580 return lh1->cookie == lh2->cookie;
583 static inline void lustre_handle_copy(struct lustre_handle *tgt,
584 const struct lustre_handle *src)
586 tgt->cookie = src->cookie;
589 /* lustre_msg struct magic. DON'T use swabbed values of MAGIC as magic! */
590 enum lustre_msg_magic {
591 LUSTRE_MSG_MAGIC_V2 = 0x0BD00BD3,
592 LUSTRE_MSG_MAGIC_V2_SWABBED = 0xD30BD00B,
593 LUSTRE_MSG_MAGIC = LUSTRE_MSG_MAGIC_V2
596 /* flags for lm_flags */
598 MSGHDR_AT_SUPPORT = 0x1, /* adaptive timeouts, lm_cksum valid
599 * in early reply messages */
600 MSGHDR_CKSUM_INCOMPAT18 = 0x2, /* compat for 1.8, needs to be set well
601 * beyond 2.8.0 for compatibility */
604 #define lustre_msg lustre_msg_v2
605 /* we depend on this structure to be 8-byte aligned */
606 /* this type is only endian-adjusted in lustre_unpack_msg() */
607 struct lustre_msg_v2 {
608 __u32 lm_bufcount; /* number of buffers in lm_buflens[] */
609 __u32 lm_secflvr; /* 0 = no crypto, or sptlrpc security flavour */
610 __u32 lm_magic; /* RPC version magic = LUSTRE_MSG_MAGIC_V2 */
611 __u32 lm_repsize; /* size of preallocated reply buffer */
612 __u32 lm_cksum; /* CRC32 of ptlrpc_body early reply messages */
613 __u32 lm_flags; /* enum lustre_msghdr MSGHDR_* flags */
614 __u32 lm_padding_2; /* unused */
615 __u32 lm_padding_3; /* unused */
616 __u32 lm_buflens[0]; /* length of additional buffers in bytes,
617 * padded to a multiple of 8 bytes. */
619 * message buffers are packed after padded lm_buflens[] array,
620 * padded to a multiple of 8 bytes each to align contents.
624 /* ptlrpc_body packet pb_types */
625 #define PTL_RPC_MSG_REQUEST 4711 /* normal RPC request message */
626 #define PTL_RPC_MSG_ERR 4712 /* error reply if request unprocessed */
627 #define PTL_RPC_MSG_REPLY 4713 /* normal RPC reply message */
629 /* ptlrpc_body pb_version ((target_version << 16) | rpc_version) */
630 enum lustre_msg_version {
631 PTLRPC_MSG_VERSION = 0x00000003,
632 LUSTRE_VERSION_MASK = 0xffff0000,
633 LUSTRE_OBD_VERSION = 0x00010000,
634 LUSTRE_MDS_VERSION = 0x00020000,
635 LUSTRE_OST_VERSION = 0x00030000,
636 LUSTRE_DLM_VERSION = 0x00040000,
637 LUSTRE_LOG_VERSION = 0x00050000,
638 LUSTRE_MGS_VERSION = 0x00060000,
641 /* pb_flags that apply to all request messages */
642 /* #define MSG_LAST_REPLAY 0x0001 obsolete 2.0 => {REQ,LOCK}_REPLAY_DONE */
643 #define MSG_RESENT 0x0002 /* was previously sent, no reply seen */
644 #define MSG_REPLAY 0x0004 /* was processed, got reply, recovery */
645 /* #define MSG_AT_SUPPORT 0x0008 obsolete since 1.5, AT always enabled */
646 /* #define MSG_DELAY_REPLAY 0x0010 obsolete since 2.0 */
647 /* #define MSG_VERSION_REPLAY 0x0020 obsolete since 1.8.2, VBR always on */
648 #define MSG_REQ_REPLAY_DONE 0x0040 /* request replay over, locks next */
649 #define MSG_LOCK_REPLAY_DONE 0x0080 /* lock replay over, client done */
651 /* pb_op_flags for connect opcodes: MDS_CONNECT, OST_CONNECT, MGS_CONNECT */
652 #define MSG_CONNECT_RECOVERING 0x00000001 /* target is in recovery */
653 #define MSG_CONNECT_RECONNECT 0x00000002 /* tgt already has client import */
654 #define MSG_CONNECT_REPLAYABLE 0x00000004 /* target supports RPC replay */
655 /* #define MSG_CONNECT_PEER 0x00000008 obsolete since 1.2, removed in 1.5 */
656 #define MSG_CONNECT_LIBCLIENT 0x00000010 /* obsolete since 2.3, removed 2.6 */
657 #define MSG_CONNECT_INITIAL 0x00000020 /* first client connection attempt */
658 /* #define MSG_CONNECT_ASYNC 0x00000040 obsolete since 1.5 */
659 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
660 #define MSG_CONNECT_TRANSNO 0x00000100 /* client sent transno in replay */
662 /* number of previous object versions in pb_pre_versions[] */
663 #define PTLRPC_NUM_VERSIONS 4
664 /* without gss, ptlrpc_body is put at the first buffer. */
665 struct ptlrpc_body_v3 {
666 struct lustre_handle pb_handle;
667 __u32 pb_type; /* request/reply/err type: PTL_RPC_MSG_* */
668 __u32 pb_version; /* LUSTRE_*_VERSION | PTLRPC_MSG_VERSION */
669 __u32 pb_opc; /* RPC opcodes: MDS_*, OST_*, LDLM_, ... */
670 __u32 pb_status; /* negative Linux x86 error number */
671 __u64 pb_last_xid; /* highest replied XID w/o lower unreplied XID*/
672 __u16 pb_tag; /* multiple modifying RPCs virtual slot index */
675 __u64 pb_last_committed;/* rep: highest pb_transno committed to disk */
676 __u64 pb_transno; /* server-assigned transno for modifying RPCs */
677 __u32 pb_flags; /* req: MSG_* flags */
678 __u32 pb_op_flags; /* req: MSG_CONNECT_* flags */
679 __u32 pb_conn_cnt; /* connect instance of this client on server */
680 __u32 pb_timeout; /* req: max wait time; rep: service estimate */
681 __u32 pb_service_time; /* rep: server arrival to reply in seconds */
682 __u32 pb_limit; /* rep: dynamic DLM LRU lock count limit */
683 __u64 pb_slv; /* rep: dynamic DLM LRU server lock volume */
684 /* VBR: rep: previous pb_version(s) of objects modified by this RPC */
685 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
686 __u64 pb_mbits; /**< match bits for bulk request */
687 /* padding for future needs - fix lustre_swab_ptlrpc_body() also */
688 __u64 pb_padding64_0;
689 __u64 pb_padding64_1;
690 __u64 pb_padding64_2;
691 char pb_jobid[LUSTRE_JOBID_SIZE]; /* req: ASCII jobid from env + NUL */
693 #define ptlrpc_body ptlrpc_body_v3
695 struct ptlrpc_body_v2 {
696 struct lustre_handle pb_handle;
701 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
702 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
705 __u64 pb_last_committed;
710 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
711 __u32 pb_service_time; /* for rep, actual service time, also used for
712 net_latency of req */
715 /* VBR: pre-versions */
716 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
717 __u64 pb_mbits; /**< unused in V2 */
718 /* padding for future needs */
719 __u64 pb_padding64_0;
720 __u64 pb_padding64_1;
721 __u64 pb_padding64_2;
724 /* message body offset for lustre_msg_v2 */
725 /* ptlrpc body offset in all request/reply messages */
726 #define MSG_PTLRPC_BODY_OFF 0
728 /* normal request/reply message record offset */
729 #define REQ_REC_OFF 1
730 #define REPLY_REC_OFF 1
732 /* ldlm request message body offset */
733 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
734 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
736 /* ldlm intent lock message body offset */
737 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
738 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
740 /* ldlm reply message body offset */
741 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
742 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
744 /** only use in req->rq_{req,rep}_swab_mask */
745 #define MSG_PTLRPC_HEADER_OFF 31
748 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
749 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
750 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
751 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
752 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
753 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
754 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
755 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
756 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
757 #define OBD_CONNECT_LARGE_ACL 0x200ULL /* more than 32 ACL entries */
758 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
759 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
760 #define OBD_CONNECT_IBITS 0x1000ULL /* not checked in 2.11+ */
761 #define OBD_CONNECT_BARRIER 0x2000ULL /* write barrier */
762 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
763 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
764 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
765 * in production. Removed in
766 * 2.9. Keep this flag to
769 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
770 * never used in production.
771 * Removed in 2.9. Keep this
772 * flag to avoid reusing.
774 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
775 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
776 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
777 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
778 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
779 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
780 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
781 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
782 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
783 #define OBD_CONNECT_REAL 0x8000000ULL /* obsolete since 2.8 */
784 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
785 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
786 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
787 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
788 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
789 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
790 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
791 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
792 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
793 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
794 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
796 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
797 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
798 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
799 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
800 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
801 * RPC error properly */
802 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
803 * finer space reservation */
804 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
805 * policy and 2.x server */
806 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
807 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
808 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
809 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
810 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
811 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* improved flock deadlock detection */
812 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/* create stripe disposition*/
813 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
815 #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */
816 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */
817 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify
819 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
820 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
821 #define OBD_CONNECT_LOCKAHEAD_OLD 0x1000000000000000ULL /* Old Cray lockahead */
823 /** bulk matchbits is sent within ptlrpc_body */
824 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL
825 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
826 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
827 /* ocd_connect_flags2 flags */
828 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* set file security context at create */
829 #define OBD_CONNECT2_LOCKAHEAD 0x2ULL /* ladvise lockahead v2 */
830 #define OBD_CONNECT2_DIR_MIGRATE 0x4ULL /* migrate striped dir */
831 #define OBD_CONNECT2_SUM_STATFS 0x8ULL /* MDT return aggregated stats */
832 #define OBD_CONNECT2_FLR 0x20ULL /* FLR support */
833 #define OBD_CONNECT2_WBC_INTENTS 0x40ULL /* create/unlink/... intents for wbc, also operations under client-held parent locks */
834 #define OBD_CONNECT2_LOCK_CONVERT 0x80ULL /* IBITS lock convert support */
835 #define OBD_CONNECT2_ARCHIVE_ID_ARRAY 0x100ULL /* store HSM archive_id in array */
836 #define OBD_CONNECT2_SELINUX_POLICY 0x400ULL /* has client SELinux policy */
839 * Please DO NOT add flag values here before first ensuring that this same
840 * flag value is not in use on some other branch. Please clear any such
841 * changes with senior engineers before starting to use a new flag. Then,
842 * submit a small patch against EVERY branch that ONLY adds the new flag,
843 * updates obd_connect_names[] for lprocfs_rd_connect_flags(), adds the
844 * flag to check_obd_connect_data(), and updates wiretests accordingly, so it
845 * can be approved and landed easily to reserve the flag for future use. */
847 /* The MNE_SWAB flag is overloading the MDS_MDS bit only for the MGS
848 * connection. It is a temporary bug fix for Imperative Recovery interop
849 * between 2.2 and 2.3 x86/ppc nodes, and can be removed when interop for
850 * 2.2 clients/servers is no longer needed. LU-1252/LU-1644. */
851 #define OBD_CONNECT_MNE_SWAB OBD_CONNECT_MDS_MDS
853 #define OCD_HAS_FLAG(ocd, flg) \
854 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
857 #ifdef HAVE_LRU_RESIZE_SUPPORT
858 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
860 #define LRU_RESIZE_CONNECT_FLAG 0
863 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
864 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
865 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
866 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
867 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
868 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
869 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
870 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
871 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
872 OBD_CONNECT_EINPROGRESS | \
873 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
874 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
875 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
876 OBD_CONNECT_FLOCK_DEAD | \
877 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
878 OBD_CONNECT_OPEN_BY_FID | \
879 OBD_CONNECT_DIR_STRIPE | OBD_CONNECT_GRANT | \
880 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_SRVLOCK | \
881 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_CKSUM | \
882 OBD_CONNECT_MULTIMODRPCS | \
883 OBD_CONNECT_SUBTREE | OBD_CONNECT_LARGE_ACL | \
884 OBD_CONNECT_GRANT_PARAM | \
885 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
887 #define MDT_CONNECT_SUPPORTED2 (OBD_CONNECT2_FILE_SECCTX | OBD_CONNECT2_FLR | \
888 OBD_CONNECT2_SUM_STATFS | \
889 OBD_CONNECT2_LOCK_CONVERT | \
890 OBD_CONNECT2_DIR_MIGRATE | \
891 OBD_CONNECT2_ARCHIVE_ID_ARRAY | \
892 OBD_CONNECT2_SELINUX_POLICY)
894 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
895 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
896 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
897 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
898 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
899 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
900 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
901 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
902 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
903 OBD_CONNECT_MAX_EASIZE | \
904 OBD_CONNECT_EINPROGRESS | \
905 OBD_CONNECT_JOBSTATS | \
906 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
907 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
908 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
909 OBD_CONNECT_BULK_MBITS | \
910 OBD_CONNECT_GRANT_PARAM | \
911 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
913 #define OST_CONNECT_SUPPORTED2 OBD_CONNECT2_LOCKAHEAD
915 #define ECHO_CONNECT_SUPPORTED (OBD_CONNECT_FID)
916 #define ECHO_CONNECT_SUPPORTED2 0
918 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
919 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
920 OBD_CONNECT_MNE_SWAB | OBD_CONNECT_PINGLESS |\
921 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_BARRIER)
923 #define MGS_CONNECT_SUPPORTED2 0
925 /* Features required for this version of the client to work with server */
926 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_FID | \
927 OBD_CONNECT_ATTRFID | \
930 /* This structure is used for both request and reply.
932 * If we eventually have separate connect data for different types, which we
933 * almost certainly will, then perhaps we stick a union in here. */
934 struct obd_connect_data {
935 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
936 __u32 ocd_version; /* lustre release version number */
937 __u32 ocd_grant; /* initial cache grant amount (bytes) */
938 __u32 ocd_index; /* LOV index to connect to */
939 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
940 __u64 ocd_ibits_known; /* inode bits this client understands */
941 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
942 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
943 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
944 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
945 __u64 ocd_transno; /* first transno from client to be replayed */
946 __u32 ocd_group; /* MDS group on OST */
947 __u32 ocd_cksum_types; /* supported checksum algorithms */
948 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
949 __u32 ocd_instance; /* instance # of this target */
950 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
951 /* Fields after ocd_maxbytes are only accessible by the receiver
952 * if the corresponding flag in ocd_connect_flags is set. Accessing
953 * any field after ocd_maxbytes on the receiver without a valid flag
954 * may result in out-of-bound memory access and kernel oops. */
955 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
956 __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */
957 __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */
958 __u64 ocd_connect_flags2;
959 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
960 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
961 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
962 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
963 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
964 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
965 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
966 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
967 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
968 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
969 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
970 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
971 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
974 * Please DO NOT use any fields here before first ensuring that this same
975 * field is not in use on some other branch. Please clear any such changes
976 * with senior engineers before starting to use a new field. Then, submit
977 * a small patch against EVERY branch that ONLY adds the new field along with
978 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
979 * reserve the flag for future use. */
982 * Supported checksum algorithms. Up to 32 checksum types are supported.
983 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
984 * Please update DECLARE_CKSUM_NAME in obd_cksum.h when adding a new
985 * algorithm and also the OBD_FL_CKSUM* flags, OBD_CKSUM_ALL flag,
986 * OBD_FL_CKSUM_ALL flag and potentially OBD_CKSUM_T10_ALL flag.
989 OBD_CKSUM_CRC32 = 0x00000001,
990 OBD_CKSUM_ADLER = 0x00000002,
991 OBD_CKSUM_CRC32C = 0x00000004,
992 OBD_CKSUM_RESERVED = 0x00000008,
993 OBD_CKSUM_T10IP512 = 0x00000010,
994 OBD_CKSUM_T10IP4K = 0x00000020,
995 OBD_CKSUM_T10CRC512 = 0x00000040,
996 OBD_CKSUM_T10CRC4K = 0x00000080,
999 #define OBD_CKSUM_T10_ALL (OBD_CKSUM_T10IP512 | OBD_CKSUM_T10IP4K | \
1000 OBD_CKSUM_T10CRC512 | OBD_CKSUM_T10CRC4K)
1002 #define OBD_CKSUM_ALL (OBD_CKSUM_CRC32 | OBD_CKSUM_ADLER | OBD_CKSUM_CRC32C | \
1006 * The default checksum algorithm used on top of T10PI GRD tags for RPC.
1007 * Considering that the checksum-of-checksums is only computing CRC32 on a
1008 * 4KB chunk of GRD tags for a 1MB RPC for 512B sectors, or 16KB of GRD
1009 * tags for 16MB of 4KB sectors, this is only 1/256 or 1/1024 of the
1010 * total data being checksummed, so the checksum type used here should not
1011 * affect overall system performance noticeably.
1013 #define OBD_CKSUM_T10_TOP OBD_CKSUM_ADLER
1016 * OST requests: OBDO & OBD request records
1021 OST_REPLY = 0, /* reply ? */
1037 OST_QUOTACHECK = 18, /* not used since 2.4 */
1039 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1041 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
1043 #define OST_FIRST_OPC OST_REPLY
1046 OBD_FL_INLINEDATA = 0x00000001,
1047 OBD_FL_OBDMDEXISTS = 0x00000002,
1048 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1049 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1050 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1051 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1052 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1053 OBD_FL_NO_PRJQUOTA = 0x00000080, /* the object's project is over
1055 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1056 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1057 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1058 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1059 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1060 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1061 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1062 OBD_FL_CKSUM_T10IP512 = 0x00005000, /* T10PI IP cksum, 512B sector */
1063 OBD_FL_CKSUM_T10IP4K = 0x00006000, /* T10PI IP cksum, 4KB sector */
1064 OBD_FL_CKSUM_T10CRC512 = 0x00007000, /* T10PI CRC cksum, 512B sector */
1065 OBD_FL_CKSUM_T10CRC4K = 0x00008000, /* T10PI CRC cksum, 4KB sector */
1066 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1067 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1068 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1069 * XXX: obsoleted - reserved for old
1070 * clients prior than 2.2 */
1071 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1072 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1073 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1074 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1075 /* OBD_FL_LOCAL_MASK = 0xF0000000, was local-only flags until 2.10 */
1078 * Note that while the original checksum values were separate bits,
1079 * in 2.x we can actually allow all values from 1-31. T10-PI checksum
1080 * types already use values which are not separate bits.
1082 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1083 OBD_FL_CKSUM_CRC32C | OBD_FL_CKSUM_T10IP512 |
1084 OBD_FL_CKSUM_T10IP4K | OBD_FL_CKSUM_T10CRC512 |
1085 OBD_FL_CKSUM_T10CRC4K,
1089 * All LOV EA magics should have the same postfix, if some new version
1090 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1091 * Lustre, even though the old version system does not recognizes such
1092 * new magic, it still can distinguish the corrupted cases by checking
1093 * the magic's postfix.
1095 #define LOV_MAGIC_MAGIC 0x0BD0
1096 #define LOV_MAGIC_MASK 0xFFFF
1098 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1099 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1100 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1101 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1102 /* reserved for specifying OSTs */
1103 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1104 #define LOV_MAGIC LOV_MAGIC_V1
1105 #define LOV_MAGIC_COMP_V1 (0x0BD60000 | LOV_MAGIC_MAGIC)
1108 * magic for fully defined striping
1109 * the idea is that we should have different magics for striping "hints"
1110 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1111 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1112 * we can't just change it w/o long way preparation, but we still need a
1113 * mechanism to allow LOD to differentiate hint versus ready striping.
1114 * so, at the moment we do a trick: MDT knows what to expect from request
1115 * depending on the case (replay uses ready striping, non-replay req uses
1116 * hints), so MDT replaces magic with appropriate one and now LOD can
1117 * easily understand what's inside -bzzz
1119 * those *_DEF magics are only used on server side internally, they
1120 * won't be put on wire or disk.
1122 #define LOV_MAGIC_DEFINED 0x10000000
1123 #define LOV_MAGIC_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V1)
1124 #define LOV_MAGIC_V3_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V3)
1125 #define LOV_MAGIC_COMP_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_COMP_V1)
1127 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1128 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1130 #define lov_ost_data lov_ost_data_v1
1131 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1132 struct ost_id l_ost_oi; /* OST object ID */
1133 __u32 l_ost_gen; /* generation of this l_ost_idx */
1134 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1137 #define lov_mds_md lov_mds_md_v1
1138 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1139 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1140 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1141 struct ost_id lmm_oi; /* LOV object ID */
1142 __u32 lmm_stripe_size; /* size of stripe in bytes */
1143 /* lmm_stripe_count used to be __u32 */
1144 __u16 lmm_stripe_count; /* num stripes in use for this object */
1145 __u16 lmm_layout_gen; /* layout generation number */
1146 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1149 #define MAX_MD_SIZE_OLD (sizeof(struct lov_mds_md) + \
1150 4 * sizeof(struct lov_ost_data))
1151 #define MAX_MD_SIZE (sizeof(struct lov_comp_md_v1) + \
1152 4 * (sizeof(struct lov_comp_md_entry_v1) + \
1154 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1156 /* This is the default MDT reply size allocated, should the striping be bigger,
1157 * it will be reallocated in mdt_fix_reply.
1158 * 100 stripes is a bit less than 2.5k of data */
1159 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1160 100 * sizeof(struct lov_ost_data))
1162 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1163 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1164 #define XATTR_USER_PREFIX "user."
1165 #define XATTR_TRUSTED_PREFIX "trusted."
1166 #define XATTR_SECURITY_PREFIX "security."
1168 #define XATTR_NAME_SOM "trusted.som"
1169 #define XATTR_NAME_LOV "trusted.lov"
1170 #define XATTR_NAME_LMA "trusted.lma"
1171 #define XATTR_NAME_LMV "trusted.lmv"
1172 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1173 #define XATTR_NAME_LINK "trusted.link"
1174 #define XATTR_NAME_FID "trusted.fid"
1175 #define XATTR_NAME_VERSION "trusted.version"
1176 #define XATTR_NAME_SOM "trusted.som"
1177 #define XATTR_NAME_HSM "trusted.hsm"
1178 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1179 #define XATTR_NAME_DUMMY "trusted.dummy"
1181 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1182 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1184 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1185 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1186 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1187 struct ost_id lmm_oi; /* LOV object ID */
1188 __u32 lmm_stripe_size; /* size of stripe in bytes */
1189 /* lmm_stripe_count used to be __u32 */
1190 __u16 lmm_stripe_count; /* num stripes in use for this object */
1191 __u16 lmm_layout_gen; /* layout generation number */
1192 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1193 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1196 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1198 if (stripes == (__u16)-1)
1201 if (lmm_magic == LOV_MAGIC_V3)
1202 return sizeof(struct lov_mds_md_v3) +
1203 stripes * sizeof(struct lov_ost_data_v1);
1205 return sizeof(struct lov_mds_md_v1) +
1206 stripes * sizeof(struct lov_ost_data_v1);
1210 lov_mds_md_max_stripe_count(size_t buf_size, __u32 lmm_magic)
1212 switch (lmm_magic) {
1213 case LOV_MAGIC_V1: {
1214 struct lov_mds_md_v1 lmm;
1216 if (buf_size < sizeof(lmm))
1219 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1221 case LOV_MAGIC_V3: {
1222 struct lov_mds_md_v3 lmm;
1224 if (buf_size < sizeof(lmm))
1227 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1234 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1235 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1236 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1237 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1238 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1239 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1240 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1241 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1242 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1243 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1244 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1245 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1246 #define OBD_MD_DOM_SIZE (0X00001000ULL) /* Data-on-MDT component size */
1247 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1248 #define OBD_MD_FLPARENT (0x00004000ULL) /* parent FID */
1249 #define OBD_MD_LAYOUT_VERSION (0x00008000ULL) /* OST object layout version */
1250 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1251 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1252 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1253 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1254 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1255 /* OBD_MD_FLQOS (0x00200000ULL) has never been used */
1256 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1257 #define OBD_MD_FLPRJQUOTA (0x00400000ULL) /* over quota flags sent from ost */
1258 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1259 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1260 /* OBD_MD_FLEPOCH (0x04000000ULL) obsolete 2.7.50 */
1261 /* ->mds if epoch opens or closes */
1262 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1263 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1264 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1265 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1266 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1268 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1269 /* OBD_MD_REINT (0x0000000200000000ULL) obsolete 1.8 */
1270 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1271 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1273 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1274 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1275 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1276 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1277 #define OBD_MD_FLAGSTATFS (0x0000010000000000ULL) /* aggregated statfs */
1278 /* OBD_MD_FLMDSCAPA (0x0000020000000000ULL) obsolete 2.7.54 */
1279 /* OBD_MD_FLOSSCAPA (0x0000040000000000ULL) obsolete 2.7.54 */
1280 /* OBD_MD_FLCKSPLIT (0x0000080000000000ULL) obsolete 2.3.58*/
1281 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1282 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1283 * under lock; for xattr
1284 * requests means the
1285 * client holds the lock */
1286 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1288 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1289 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1292 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1293 #define OBD_MD_FLOSTLAYOUT (0x0080000000000000ULL) /* contain ost_layout */
1294 #define OBD_MD_FLPROJID (0x0100000000000000ULL) /* project ID */
1295 #define OBD_MD_SECCTX (0x0200000000000000ULL) /* embed security xattr */
1297 #define OBD_MD_FLALLQUOTA (OBD_MD_FLUSRQUOTA | \
1298 OBD_MD_FLGRPQUOTA | \
1301 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1302 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1303 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1304 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1305 OBD_MD_FLPARENT | OBD_MD_FLRDEV | OBD_MD_FLGROUP | \
1308 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1310 /* don't forget obdo_fid which is way down at the bottom so it can
1311 * come after the definition of llog_cookie */
1315 HSS_CLEARMASK = 0x02,
1316 HSS_ARCHIVE_ID = 0x04,
1319 struct hsm_state_set {
1321 __u32 hss_archive_id;
1323 __u64 hss_clearmask;
1326 /* ost_body.data values for OST_BRW */
1328 #define OBD_BRW_READ 0x01
1329 #define OBD_BRW_WRITE 0x02
1330 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1331 #define OBD_BRW_NDELAY 0x04 /* Non-delay RPC should be issued for
1332 * this page. Non-delay RPCs have bit
1333 * rq_no_delay set. */
1334 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1335 * transfer and is not accounted in
1337 #define OBD_BRW_CHECK 0x10
1338 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1339 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1340 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1341 #define OBD_BRW_NOQUOTA 0x100
1342 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1343 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1344 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1345 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1346 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1347 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1348 * that the client is running low on
1349 * space for unstable pages; asking
1350 * it to sync quickly */
1351 #define OBD_BRW_OVER_PRJQUOTA 0x8000 /* Running out of project quota */
1353 #define OBD_BRW_OVER_ALLQUOTA (OBD_BRW_OVER_USRQUOTA | \
1354 OBD_BRW_OVER_GRPQUOTA | \
1355 OBD_BRW_OVER_PRJQUOTA)
1357 #define OBD_BRW_LOCAL1 0x80000000UL /*
1358 * osd-ldiskfs internal,
1359 * page mapped to real block
1362 #define OBD_BRW_LOCALS (OBD_BRW_LOCAL1)
1364 #define OBD_MAX_GRANT 0x7fffffffUL /* Max grant allowed to one client: 2 GiB */
1366 #define OBD_OBJECT_EOF LUSTRE_EOF
1368 #define OST_MIN_PRECREATE 32
1369 #define OST_MAX_PRECREATE 20000
1372 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1373 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1374 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1375 * high 16 bits in 2.4 and later */
1376 __u32 ioo_bufcnt; /* number of niobufs for this object */
1379 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1380 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1381 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. */
1382 #define IOOBJ_MAX_BRW_BITS 16
1383 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1384 #define ioobj_max_brw_set(ioo, num) \
1385 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1387 /* multiple of 8 bytes => can array */
1388 struct niobuf_remote {
1394 /* lock value block communicated between the filter and llite */
1396 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1397 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1398 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1399 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1400 #define OST_LVB_IS_ERR(blocks) \
1401 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1402 #define OST_LVB_SET_ERR(blocks, rc) \
1403 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1404 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1427 * lquota data structures
1430 #ifndef QUOTABLOCK_BITS
1431 # define QUOTABLOCK_BITS LUSTRE_QUOTABLOCK_BITS
1434 #ifndef QUOTABLOCK_SIZE
1435 # define QUOTABLOCK_SIZE LUSTRE_QUOTABLOCK_SIZE
1439 # define toqb lustre_stoqb
1442 /* The lquota_id structure is an union of all the possible identifier types that
1443 * can be used with quota, this includes:
1446 * - a FID which can be used for per-directory quota in the future */
1448 struct lu_fid qid_fid; /* FID for per-directory quota */
1449 __u64 qid_uid; /* user identifier */
1450 __u64 qid_gid; /* group identifier */
1451 __u64 qid_projid; /* project identifier */
1454 /* quotactl management */
1455 struct obd_quotactl {
1457 __u32 qc_type; /* see Q_* flag below */
1460 struct obd_dqinfo qc_dqinfo;
1461 struct obd_dqblk qc_dqblk;
1464 #define Q_COPY(out, in, member) (out)->member = (in)->member
1466 #define QCTL_COPY(out, in) \
1468 Q_COPY(out, in, qc_cmd); \
1469 Q_COPY(out, in, qc_type); \
1470 Q_COPY(out, in, qc_id); \
1471 Q_COPY(out, in, qc_stat); \
1472 Q_COPY(out, in, qc_dqinfo); \
1473 Q_COPY(out, in, qc_dqblk); \
1476 /* Body of quota request used for quota acquire/release RPCs between quota
1477 * master (aka QMT) and slaves (ak QSD). */
1479 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1480 * and type (data or metadata) as well as
1481 * the quota type (user or group). */
1482 union lquota_id qb_id; /* uid or gid or directory FID */
1483 __u32 qb_flags; /* see below */
1485 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1486 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1487 __u64 qb_slv_ver; /* slave index file version */
1488 struct lustre_handle qb_lockh; /* per-ID lock handle */
1489 struct lustre_handle qb_glb_lockh; /* global lock handle */
1490 __u64 qb_padding1[4];
1493 /* When the quota_body is used in the reply of quota global intent
1494 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1495 #define qb_slv_fid qb_fid
1496 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1498 #define qb_qunit qb_usage
1500 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1501 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1502 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1503 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1505 /* Quota types currently supported */
1507 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1508 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1509 LQUOTA_TYPE_PRJ = 0x02, /* maps to PRJQUOTA */
1513 /* There are 2 different resource types on which a quota limit can be enforced:
1514 * - inodes on the MDTs
1515 * - blocks on the OSTs */
1517 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1518 LQUOTA_RES_DT = 0x02,
1520 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1522 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1525 * Space accounting support
1526 * Format of an accounting record, providing disk usage information for a given
1529 struct lquota_acct_rec { /* 16 bytes */
1530 __u64 bspace; /* current space in use */
1531 __u64 ispace; /* current # inodes in use */
1535 * Global quota index support
1536 * Format of a global record, providing global quota settings for a given quota
1539 struct lquota_glb_rec { /* 32 bytes */
1540 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1541 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1542 __u64 qbr_time; /* grace time, in seconds */
1543 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1548 * Slave index support
1549 * Format of a slave record, recording how much space is granted to a given
1552 struct lquota_slv_rec { /* 8 bytes */
1553 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1554 * in #inodes or kbytes */
1557 /* Data structures associated with the quota locks */
1559 /* Glimpse descriptor used for the index & per-ID quota locks */
1560 struct ldlm_gl_lquota_desc {
1561 union lquota_id gl_id; /* quota ID subject to the glimpse */
1562 __u64 gl_flags; /* see LQUOTA_FL* below */
1563 __u64 gl_ver; /* new index version */
1564 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1565 __u64 gl_softlimit; /* new softlimit */
1569 #define gl_qunit gl_hardlimit /* current qunit value used when
1570 * glimpsing per-ID quota locks */
1572 /* quota glimpse flags */
1573 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1575 /* LVB used with quota (global and per-ID) locks */
1577 __u64 lvb_flags; /* see LQUOTA_FL* above */
1578 __u64 lvb_id_may_rel; /* space that might be released later */
1579 __u64 lvb_id_rel; /* space released by the slave for this ID */
1580 __u64 lvb_id_qunit; /* current qunit value */
1584 /* LVB used with global quota lock */
1585 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1593 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1602 MDS_GETATTR_NAME = 34,
1607 MDS_DISCONNECT = 39,
1610 MDS_PIN = 42, /* obsolete, never used in a release */
1611 MDS_UNPIN = 43, /* obsolete, never used in a release */
1613 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1615 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1618 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1620 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1622 MDS_HSM_STATE_GET = 54,
1623 MDS_HSM_STATE_SET = 55,
1624 MDS_HSM_ACTION = 56,
1625 MDS_HSM_PROGRESS = 57,
1626 MDS_HSM_REQUEST = 58,
1627 MDS_HSM_CT_REGISTER = 59,
1628 MDS_HSM_CT_UNREGISTER = 60,
1629 MDS_SWAP_LAYOUTS = 61,
1633 #define MDS_FIRST_OPC MDS_GETATTR
1636 /* opcodes for object update */
1642 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1662 /* the disposition of the intent outlines what was executed */
1663 #define DISP_IT_EXECD 0x00000001
1664 #define DISP_LOOKUP_EXECD 0x00000002
1665 #define DISP_LOOKUP_NEG 0x00000004
1666 #define DISP_LOOKUP_POS 0x00000008
1667 #define DISP_OPEN_CREATE 0x00000010
1668 #define DISP_OPEN_OPEN 0x00000020
1669 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1670 #define DISP_ENQ_OPEN_REF 0x00800000
1671 #define DISP_ENQ_CREATE_REF 0x01000000
1672 #define DISP_OPEN_LOCK 0x02000000
1673 #define DISP_OPEN_LEASE 0x04000000
1674 #define DISP_OPEN_STRIPE 0x08000000
1675 #define DISP_OPEN_DENY 0x10000000
1677 /* INODE LOCK PARTS */
1678 #define MDS_INODELOCK_LOOKUP 0x000001 /* For namespace, dentry etc, and also
1679 * was used to protect permission (mode,
1680 * owner, group etc) before 2.4. */
1681 #define MDS_INODELOCK_UPDATE 0x000002 /* size, links, timestamps */
1682 #define MDS_INODELOCK_OPEN 0x000004 /* For opened files */
1683 #define MDS_INODELOCK_LAYOUT 0x000008 /* for layout */
1685 /* The PERM bit is added int 2.4, and it is used to protect permission(mode,
1686 * owner, group, acl etc), so to separate the permission from LOOKUP lock.
1687 * Because for remote directories(in DNE), these locks will be granted by
1688 * different MDTs(different ldlm namespace).
1690 * For local directory, MDT will always grant UPDATE_LOCK|PERM_LOCK together.
1691 * For Remote directory, the master MDT, where the remote directory is, will
1692 * grant UPDATE_LOCK|PERM_LOCK, and the remote MDT, where the name entry is,
1693 * will grant LOOKUP_LOCK. */
1694 #define MDS_INODELOCK_PERM 0x000010
1695 #define MDS_INODELOCK_XATTR 0x000020 /* extended attributes */
1696 #define MDS_INODELOCK_DOM 0x000040 /* Data for data-on-mdt files */
1698 #define MDS_INODELOCK_MAXSHIFT 6
1699 /* This FULL lock is useful to take on unlink sort of operations */
1700 #define MDS_INODELOCK_FULL ((1<<(MDS_INODELOCK_MAXSHIFT+1))-1)
1701 /* DOM lock shouldn't be canceled early, use this macro for ELC */
1702 #define MDS_INODELOCK_ELC (MDS_INODELOCK_FULL & ~MDS_INODELOCK_DOM)
1704 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1705 * but was moved into name[1] along with the OID to avoid consuming the
1706 * name[2,3] fields that need to be used for the quota id (also a FID). */
1708 LUSTRE_RES_ID_SEQ_OFF = 0,
1709 LUSTRE_RES_ID_VER_OID_OFF = 1,
1710 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1711 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1712 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1713 LUSTRE_RES_ID_HSH_OFF = 3
1716 #define MDS_STATUS_CONN 1
1717 #define MDS_STATUS_LOV 2
1720 /* these should be identical to their EXT4_*_FL counterparts, they are
1721 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1722 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1723 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1724 LUSTRE_APPEND_FL = 0x00000020, /* file writes may only append */
1725 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1726 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1727 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1728 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1729 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1730 LUSTRE_DIRECTIO_FL = 0x00100000, /* Use direct i/o */
1731 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1732 LUSTRE_PROJINHERIT_FL = 0x20000000, /* Create with parents projid */
1734 /* These flags will not be identical to any EXT4_*_FL counterparts,
1735 * and only reserved for lustre purpose. Note: these flags might
1736 * be conflict with some of EXT4 flags, so
1737 * 1. these conflict flags needs to be removed when the flag is
1738 * wired by la_flags see osd_attr_get().
1739 * 2. If these flags needs to be stored into inode, they will be
1740 * stored in LMA. see LMAI_XXXX */
1741 LUSTRE_ORPHAN_FL = 0x00002000,
1742 LUSTRE_SET_SYNC_FL = 0x00040000, /* Synchronous setattr on OSTs */
1744 LUSTRE_LMA_FL_MASKS = LUSTRE_ORPHAN_FL,
1747 #ifndef FS_XFLAG_SYNC
1748 #define FS_XFLAG_SYNC 0x00000020 /* all writes synchronous */
1750 #ifndef FS_XFLAG_NOATIME
1751 #define FS_XFLAG_NOATIME 0x00000040 /* do not update access time */
1753 #ifndef FS_XFLAG_IMMUTABLE
1754 #define FS_XFLAG_IMMUTABLE 0x00000008 /* file cannot be modified */
1756 #ifndef FS_XFLAG_APPEND
1757 #define FS_XFLAG_APPEND 0x00000010 /* all writes append */
1759 #ifndef FS_XFLAG_PROJINHERIT
1760 #define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
1764 /* Convert wire LUSTRE_*_FL to corresponding client local VFS S_* values
1765 * for the client inode i_flags. The LUSTRE_*_FL are the Lustre wire
1766 * protocol equivalents of LDISKFS_*_FL values stored on disk, while
1767 * the S_* flags are kernel-internal values that change between kernel
1768 * versions. These flags are set/cleared via FSFILT_IOC_{GET,SET}_FLAGS.
1769 * See b=16526 for a full history. */
1770 static inline int ll_ext_to_inode_flags(int flags)
1772 return (((flags & LUSTRE_SYNC_FL) ? S_SYNC : 0) |
1773 ((flags & LUSTRE_NOATIME_FL) ? S_NOATIME : 0) |
1774 ((flags & LUSTRE_APPEND_FL) ? S_APPEND : 0) |
1775 #if defined(S_DIRSYNC)
1776 ((flags & LUSTRE_DIRSYNC_FL) ? S_DIRSYNC : 0) |
1778 ((flags & LUSTRE_IMMUTABLE_FL) ? S_IMMUTABLE : 0));
1781 static inline int ll_inode_to_ext_flags(int iflags)
1783 return (((iflags & S_SYNC) ? LUSTRE_SYNC_FL : 0) |
1784 ((iflags & S_NOATIME) ? LUSTRE_NOATIME_FL : 0) |
1785 ((iflags & S_APPEND) ? LUSTRE_APPEND_FL : 0) |
1786 #if defined(S_DIRSYNC)
1787 ((iflags & S_DIRSYNC) ? LUSTRE_DIRSYNC_FL : 0) |
1789 ((iflags & S_IMMUTABLE) ? LUSTRE_IMMUTABLE_FL : 0));
1792 static inline int ll_xflags_to_inode_flags(int xflags)
1794 return ((xflags & FS_XFLAG_SYNC) ? S_SYNC : 0) |
1795 ((xflags & FS_XFLAG_NOATIME) ? S_NOATIME : 0) |
1796 ((xflags & FS_XFLAG_APPEND) ? S_APPEND : 0) |
1797 ((xflags & FS_XFLAG_IMMUTABLE) ? S_IMMUTABLE : 0);
1800 static inline int ll_inode_flags_to_xflags(int flags)
1802 return ((flags & S_SYNC) ? FS_XFLAG_SYNC : 0) |
1803 ((flags & S_NOATIME) ? FS_XFLAG_NOATIME : 0) |
1804 ((flags & S_APPEND) ? FS_XFLAG_APPEND : 0) |
1805 ((flags & S_IMMUTABLE) ? FS_XFLAG_IMMUTABLE : 0);
1809 /* 64 possible states */
1810 enum md_transient_state {
1811 MS_RESTORE = (1 << 0), /* restore is running */
1815 struct lu_fid mbo_fid1;
1816 struct lu_fid mbo_fid2;
1817 struct lustre_handle mbo_open_handle;
1819 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1823 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1824 __u64 mbo_version; /* was mbo_ioepoch before 2.11 */
1825 __u64 mbo_t_state; /* transient file state defined in
1826 * enum md_transient_state
1827 * was "ino" until 2.4.0 */
1830 __u32 mbo_capability;
1834 __u32 mbo_flags; /* LUSTRE_*_FL file attributes */
1836 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1837 __u32 mbo_layout_gen; /* was "generation" until 2.4.0 */
1839 __u32 mbo_eadatasize;
1841 __u32 mbo_max_mdsize;
1842 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1843 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1844 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1846 __u64 mbo_dom_size; /* size of DOM component */
1847 __u64 mbo_dom_blocks; /* blocks consumed by DOM component */
1848 __u64 mbo_padding_8; /* also fix lustre_swab_mdt_body */
1849 __u64 mbo_padding_9;
1850 __u64 mbo_padding_10;
1853 struct mdt_ioepoch {
1854 struct lustre_handle mio_open_handle;
1855 __u64 mio_unused1; /* was ioepoch */
1856 __u32 mio_unused2; /* was flags */
1860 /* permissions for md_perm.mp_perm */
1862 CFS_SETUID_PERM = 0x01,
1863 CFS_SETGID_PERM = 0x02,
1864 CFS_SETGRP_PERM = 0x04,
1867 struct mdt_rec_setattr {
1877 __u32 sa_padding_1_h;
1878 struct lu_fid sa_fid;
1887 __u32 sa_attr_flags;
1889 __u32 sa_bias; /* some operation flags */
1896 * Attribute flags used in mdt_rec_setattr::sa_valid.
1897 * The kernel's #defines for ATTR_* should not be used over the network
1898 * since the client and MDS may run different kernels (see bug 13828)
1899 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1901 #define MDS_ATTR_MODE 0x1ULL /* = 1 */
1902 #define MDS_ATTR_UID 0x2ULL /* = 2 */
1903 #define MDS_ATTR_GID 0x4ULL /* = 4 */
1904 #define MDS_ATTR_SIZE 0x8ULL /* = 8 */
1905 #define MDS_ATTR_ATIME 0x10ULL /* = 16 */
1906 #define MDS_ATTR_MTIME 0x20ULL /* = 32 */
1907 #define MDS_ATTR_CTIME 0x40ULL /* = 64 */
1908 #define MDS_ATTR_ATIME_SET 0x80ULL /* = 128 */
1909 #define MDS_ATTR_MTIME_SET 0x100ULL /* = 256 */
1910 #define MDS_ATTR_FORCE 0x200ULL /* = 512, Not a change, but a change it */
1911 #define MDS_ATTR_ATTR_FLAG 0x400ULL /* = 1024 */
1912 #define MDS_ATTR_KILL_SUID 0x800ULL /* = 2048 */
1913 #define MDS_ATTR_KILL_SGID 0x1000ULL /* = 4096 */
1914 #define MDS_ATTR_CTIME_SET 0x2000ULL /* = 8192 */
1915 #define MDS_ATTR_FROM_OPEN 0x4000ULL /* = 16384, called from open path, ie O_TRUNC */
1916 #define MDS_ATTR_BLOCKS 0x8000ULL /* = 32768 */
1917 #define MDS_ATTR_PROJID 0x10000ULL /* = 65536 */
1918 #define MDS_ATTR_LSIZE 0x20000ULL /* = 131072 */
1919 #define MDS_ATTR_LBLOCKS 0x40000ULL /* = 262144 */
1920 #define MDS_ATTR_OVERRIDE 0x2000000ULL /* = 33554432 */
1923 /* MDS_CHECK_SPLIT = 1 << 0, obsolete before 2.3.58 */
1924 MDS_CROSS_REF = 1 << 1,
1925 /* MDS_VTX_BYPASS = 1 << 2, obsolete since 2.3.54 */
1926 MDS_PERM_BYPASS = 1 << 3,
1927 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1928 MDS_QUOTA_IGNORE = 1 << 5,
1929 /* MDS_CLOSE_CLEANUP = 1 << 6, obsolete since 2.3.51 */
1930 MDS_KEEP_ORPHAN = 1 << 7,
1931 MDS_RECOV_OPEN = 1 << 8,
1932 MDS_DATA_MODIFIED = 1 << 9,
1933 MDS_CREATE_VOLATILE = 1 << 10,
1934 MDS_OWNEROVERRIDE = 1 << 11,
1935 MDS_HSM_RELEASE = 1 << 12,
1936 MDS_CLOSE_MIGRATE = 1 << 13,
1937 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1938 MDS_CLOSE_LAYOUT_MERGE = 1 << 15,
1939 MDS_CLOSE_RESYNC_DONE = 1 << 16,
1940 MDS_CLOSE_LAYOUT_SPLIT = 1 << 17,
1941 MDS_TRUNC_KEEP_LEASE = 1 << 18,
1944 #define MDS_CLOSE_INTENT (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP | \
1945 MDS_CLOSE_LAYOUT_MERGE | MDS_CLOSE_LAYOUT_SPLIT | \
1946 MDS_CLOSE_RESYNC_DONE)
1948 /* instance of mdt_reint_rec */
1949 struct mdt_rec_create {
1957 __u32 cr_suppgid1_h;
1959 __u32 cr_suppgid2_h;
1960 struct lu_fid cr_fid1;
1961 struct lu_fid cr_fid2;
1962 struct lustre_handle cr_open_handle_old; /* in case of open replay */
1966 __u64 cr_padding_1; /* rr_blocks */
1969 /* use of helpers set/get_mrc_cr_flags() is needed to access
1970 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
1971 * extend cr_flags size without breaking 1.8 compat */
1972 __u32 cr_flags_l; /* for use with open, low 32 bits */
1973 __u32 cr_flags_h; /* for use with open, high 32 bits */
1974 __u32 cr_umask; /* umask for create */
1975 __u32 cr_padding_4; /* rr_padding_4 */
1978 /* instance of mdt_reint_rec */
1979 struct mdt_rec_link {
1987 __u32 lk_suppgid1_h;
1989 __u32 lk_suppgid2_h;
1990 struct lu_fid lk_fid1;
1991 struct lu_fid lk_fid2;
1993 __u64 lk_padding_1; /* rr_atime */
1994 __u64 lk_padding_2; /* rr_ctime */
1995 __u64 lk_padding_3; /* rr_size */
1996 __u64 lk_padding_4; /* rr_blocks */
1998 __u32 lk_padding_5; /* rr_mode */
1999 __u32 lk_padding_6; /* rr_flags */
2000 __u32 lk_padding_7; /* rr_padding_2 */
2001 __u32 lk_padding_8; /* rr_padding_3 */
2002 __u32 lk_padding_9; /* rr_padding_4 */
2005 /* instance of mdt_reint_rec */
2006 struct mdt_rec_unlink {
2014 __u32 ul_suppgid1_h;
2016 __u32 ul_suppgid2_h;
2017 struct lu_fid ul_fid1;
2018 struct lu_fid ul_fid2;
2020 __u64 ul_padding_2; /* rr_atime */
2021 __u64 ul_padding_3; /* rr_ctime */
2022 __u64 ul_padding_4; /* rr_size */
2023 __u64 ul_padding_5; /* rr_blocks */
2026 __u32 ul_padding_6; /* rr_flags */
2027 __u32 ul_padding_7; /* rr_padding_2 */
2028 __u32 ul_padding_8; /* rr_padding_3 */
2029 __u32 ul_padding_9; /* rr_padding_4 */
2032 /* instance of mdt_reint_rec */
2033 struct mdt_rec_rename {
2041 __u32 rn_suppgid1_h;
2043 __u32 rn_suppgid2_h;
2044 struct lu_fid rn_fid1;
2045 struct lu_fid rn_fid2;
2047 __u64 rn_padding_1; /* rr_atime */
2048 __u64 rn_padding_2; /* rr_ctime */
2049 __u64 rn_padding_3; /* rr_size */
2050 __u64 rn_padding_4; /* rr_blocks */
2051 __u32 rn_bias; /* some operation flags */
2052 __u32 rn_mode; /* cross-ref rename has mode */
2053 __u32 rn_padding_5; /* rr_flags */
2054 __u32 rn_padding_6; /* rr_padding_2 */
2055 __u32 rn_padding_7; /* rr_padding_3 */
2056 __u32 rn_padding_8; /* rr_padding_4 */
2059 /* instance of mdt_reint_rec */
2060 struct mdt_rec_setxattr {
2068 __u32 sx_suppgid1_h;
2070 __u32 sx_suppgid2_h;
2071 struct lu_fid sx_fid;
2072 __u64 sx_padding_1; /* These three are rr_fid2 */
2077 __u64 sx_padding_5; /* rr_ctime */
2078 __u64 sx_padding_6; /* rr_size */
2079 __u64 sx_padding_7; /* rr_blocks */
2082 __u32 sx_padding_8; /* rr_flags */
2083 __u32 sx_padding_9; /* rr_padding_2 */
2084 __u32 sx_padding_10; /* rr_padding_3 */
2085 __u32 sx_padding_11; /* rr_padding_4 */
2088 /* instance of mdt_reint_rec
2089 * FLR: for file resync MDS_REINT_RESYNC RPC. */
2090 struct mdt_rec_resync {
2098 __u32 rs_suppgid1_h;
2100 __u32 rs_suppgid2_h;
2101 struct lu_fid rs_fid;
2102 __u8 rs_padding0[sizeof(struct lu_fid)];
2103 struct lustre_handle rs_lease_handle; /* rr_mtime */
2104 __s64 rs_padding1; /* rr_atime */
2105 __s64 rs_padding2; /* rr_ctime */
2106 __u64 rs_padding3; /* rr_size */
2107 __u64 rs_padding4; /* rr_blocks */
2109 __u32 rs_padding5; /* rr_mode */
2110 __u32 rs_padding6; /* rr_flags */
2111 __u32 rs_padding7; /* rr_flags_h */
2112 __u32 rs_padding8; /* rr_umask */
2114 __u16 rs_padding9; /* rr_padding_4 */
2118 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2119 * Do NOT change the size of various members, otherwise the value
2120 * will be broken in lustre_swab_mdt_rec_reint().
2122 * If you add new members in other mdt_reint_xxx structres and need to use the
2123 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2125 struct mdt_rec_reint {
2133 __u32 rr_suppgid1_h;
2135 __u32 rr_suppgid2_h;
2136 struct lu_fid rr_fid1;
2137 struct lu_fid rr_fid2;
2149 __u16 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2152 /* lmv structures */
2154 __u32 ld_tgt_count; /* how many MDS's */
2155 __u32 ld_active_tgt_count; /* how many active */
2156 __u32 ld_default_stripe_count; /* how many objects are used */
2157 __u32 ld_pattern; /* default hash pattern */
2158 __u64 ld_default_hash_size;
2159 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2160 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2161 __u32 ld_qos_maxage; /* in second */
2162 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2163 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2164 struct obd_uuid ld_uuid;
2167 /* LMV layout EA, and it will be stored both in master and slave object */
2168 struct lmv_mds_md_v1 {
2170 __u32 lmv_stripe_count;
2171 __u32 lmv_master_mdt_index; /* On master object, it is master
2172 * MDT index, on slave object, it
2173 * is stripe index of the slave obj */
2174 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2175 * which hash function to be used,
2176 * Note: only lower 16 bits is being
2177 * used for now. Higher 16 bits will
2178 * be used to mark the object status,
2179 * for example migrating or dead. */
2180 __u32 lmv_layout_version; /* increased each time layout changed,
2181 * by directory migration, restripe
2183 __u32 lmv_migrate_offset; /* once this is set, it means this
2184 * directory is been migrated, stripes
2185 * before this offset belong to target,
2186 * from this to source. */
2187 __u32 lmv_migrate_hash; /* hash type of source stripes of
2188 * migrating directory */
2191 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
2192 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2195 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2196 #define LMV_MAGIC LMV_MAGIC_V1
2198 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2199 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2201 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
2202 * and the higher part will be the flag to indicate the status of object,
2203 * for example the object is being migrated. And the hash function
2204 * might be interpreted differently with different flags. */
2205 #define LMV_HASH_TYPE_MASK 0x0000ffff
2207 #define LMV_HASH_FLAG_MIGRATION 0x80000000
2209 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 12, 53, 0)
2210 /* Since lustre 2.8, this flag will not be needed, instead this DEAD
2211 * and orphan flags will be stored in LMA (see LMAI_ORPHAN)
2212 * Keep this flag just for LFSCK, because it still might meet such
2213 * flag when it checks the old FS */
2214 #define LMV_HASH_FLAG_DEAD 0x40000000
2216 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
2218 /* The striped directory has ever lost its master LMV EA, then LFSCK
2219 * re-generated it. This flag is used to indicate such case. It is an
2221 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
2224 * The FNV-1a hash algorithm is as follows:
2225 * hash = FNV_offset_basis
2226 * for each octet_of_data to be hashed
2227 * hash = hash XOR octet_of_data
2228 * hash = hash × FNV_prime
2230 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2232 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2233 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2235 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2236 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2237 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, size_t size)
2239 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2240 const unsigned char *p = buf;
2243 for (i = 0; i < size; i++) {
2245 hash *= LUSTRE_FNV_1A_64_PRIME;
2253 struct lmv_mds_md_v1 lmv_md_v1;
2254 struct lmv_user_md lmv_user_md;
2257 static inline int lmv_mds_md_size(int stripe_count, unsigned int lmm_magic)
2259 switch (lmm_magic) {
2261 struct lmv_mds_md_v1 *lmm1;
2263 return sizeof(*lmm1) + stripe_count *
2264 sizeof(lmm1->lmv_stripe_fids[0]);
2271 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2273 switch (__le32_to_cpu(lmm->lmv_magic)) {
2275 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2276 case LMV_USER_MAGIC:
2277 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2283 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2285 switch (__le32_to_cpu(lmm->lmv_magic)) {
2287 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2288 case LMV_USER_MAGIC:
2289 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2299 FLD_FIRST_OPC = FLD_QUERY
2305 SEQ_FIRST_OPC = SEQ_QUERY
2309 SEQ_ALLOC_SUPER = 0,
2321 LFSCK_NOTIFY = 1101,
2324 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2328 * LOV data structures
2331 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2332 /* The size of the buffer the lov/mdc reserves for the
2333 * array of UUIDs returned by the MDS. With the current
2334 * protocol, this will limit the max number of OSTs per LOV */
2336 #define LOV_DESC_MAGIC 0xB0CCDE5C
2337 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2338 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2340 /* LOV settings descriptor (should only contain static info) */
2342 __u32 ld_tgt_count; /* how many OBD's */
2343 __u32 ld_active_tgt_count; /* how many active */
2344 __s32 ld_default_stripe_count; /* how many objects are used */
2345 __u32 ld_pattern; /* default PATTERN_RAID0 */
2346 __u64 ld_default_stripe_size; /* in bytes */
2347 __s64 ld_default_stripe_offset; /* starting OST index */
2348 __u32 ld_padding_0; /* unused */
2349 __u32 ld_qos_maxage; /* in second */
2350 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2351 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2352 struct obd_uuid ld_uuid;
2355 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2360 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2365 LDLM_BL_CALLBACK = 104,
2366 LDLM_CP_CALLBACK = 105,
2367 LDLM_GL_CALLBACK = 106,
2368 LDLM_SET_INFO = 107,
2371 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2373 #define RES_NAME_SIZE 4
2374 struct ldlm_res_id {
2375 __u64 name[RES_NAME_SIZE];
2378 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2379 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2380 (unsigned long long)(res)->lr_name.name[1], \
2381 (unsigned long long)(res)->lr_name.name[2], \
2382 (unsigned long long)(res)->lr_name.name[3]
2398 #define LCK_MODE_NUM 8
2408 #define LDLM_MIN_TYPE LDLM_PLAIN
2410 struct ldlm_extent {
2416 static inline bool ldlm_extent_equal(const struct ldlm_extent *ex1,
2417 const struct ldlm_extent *ex2)
2419 return ex1->start == ex2->start && ex1->end == ex2->end;
2422 struct ldlm_inodebits {
2425 __u64 try_bits; /* optional bits to try */
2426 __u64 cancel_bits; /* for lock convert */
2430 struct ldlm_flock_wire {
2438 /* it's important that the fields of the ldlm_extent structure match
2439 * the first fields of the ldlm_flock structure because there is only
2440 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2441 * this ever changes we will need to swab the union differently based
2442 * on the resource type. */
2444 union ldlm_wire_policy_data {
2445 struct ldlm_extent l_extent;
2446 struct ldlm_flock_wire l_flock;
2447 struct ldlm_inodebits l_inodebits;
2450 struct barrier_lvb {
2456 struct ldlm_gl_barrier_desc {
2462 union ldlm_gl_desc {
2463 struct ldlm_gl_lquota_desc lquota_desc;
2464 struct ldlm_gl_barrier_desc barrier_desc;
2467 enum ldlm_intent_flags {
2468 IT_OPEN = 0x00000001,
2469 IT_CREAT = 0x00000002,
2470 IT_OPEN_CREAT = IT_OPEN | IT_CREAT, /* To allow case label. */
2471 IT_READDIR = 0x00000004, /* Used by mdc, not put on the wire. */
2472 IT_GETATTR = 0x00000008,
2473 IT_LOOKUP = 0x00000010,
2474 /* IT_UNLINK = 0x00000020, Obsolete. */
2475 /* IT_TRUNC = 0x00000040, Obsolete. */
2476 IT_GETXATTR = 0x00000080,
2477 /* IT_EXEC = 0x00000100, Obsolete. */
2478 /* IT_PIN = 0x00000200, Obsolete. */
2479 IT_LAYOUT = 0x00000400,
2480 IT_QUOTA_DQACQ = 0x00000800,
2481 IT_QUOTA_CONN = 0x00001000,
2482 /* IT_SETXATTR = 0x00002000, Obsolete. */
2483 IT_GLIMPSE = 0x00004000,
2484 IT_BRW = 0x00008000,
2487 struct ldlm_intent {
2491 struct ldlm_resource_desc {
2492 enum ldlm_type lr_type;
2493 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2494 struct ldlm_res_id lr_name;
2497 struct ldlm_lock_desc {
2498 struct ldlm_resource_desc l_resource;
2499 enum ldlm_mode l_req_mode;
2500 enum ldlm_mode l_granted_mode;
2501 union ldlm_wire_policy_data l_policy_data;
2504 #define LDLM_LOCKREQ_HANDLES 2
2505 #define LDLM_ENQUEUE_CANCEL_OFF 1
2507 struct ldlm_request {
2508 __u32 lock_flags; /* LDLM_FL_*, see lustre_dlm_flags.h */
2509 __u32 lock_count; /* number of locks in lock_handle[] */
2510 struct ldlm_lock_desc lock_desc;/* lock descriptor */
2511 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2516 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2517 struct ldlm_lock_desc lock_desc;
2518 struct lustre_handle lock_handle;
2519 __u64 lock_policy_res1;
2520 __u64 lock_policy_res2;
2523 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2524 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2527 * Opcodes for mountconf (mgs and mgc)
2531 MGS_DISCONNECT = 251,
2532 MGS_EXCEPTION = 252, /* node died, etc. */
2533 MGS_TARGET_REG = 253, /* whenever target starts up */
2534 MGS_TARGET_DEL = 254,
2536 MGS_CONFIG_READ = 256,
2538 MGS_FIRST_OPC = MGS_CONNECT
2541 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 13, 53, 0)
2542 #define MGS_PARAM_MAXLEN 1024
2543 #define KEY_SET_INFO "set_info"
2545 struct mgs_send_param {
2546 char mgs_param[MGS_PARAM_MAXLEN];
2550 /* We pass this info to the MGS so it can write config logs */
2551 #define MTI_NAME_MAXLEN 64
2552 #define MTI_PARAM_MAXLEN 4096
2553 #define MTI_NIDS_MAX 32
2554 struct mgs_target_info {
2555 __u32 mti_lustre_ver;
2556 __u32 mti_stripe_index;
2557 __u32 mti_config_ver;
2558 __u32 mti_flags; /* LDD_F_* */
2559 __u32 mti_nid_count;
2560 __u32 mti_instance; /* Running instance of target */
2561 char mti_fsname[MTI_NAME_MAXLEN];
2562 char mti_svname[MTI_NAME_MAXLEN];
2563 char mti_uuid[sizeof(struct obd_uuid)];
2564 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t) */
2565 char mti_params[MTI_PARAM_MAXLEN];
2568 struct mgs_nidtbl_entry {
2569 __u64 mne_version; /* table version of this entry */
2570 __u32 mne_instance; /* target instance # */
2571 __u32 mne_index; /* target index */
2572 __u32 mne_length; /* length of this entry - by bytes */
2573 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2574 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2575 __u8 mne_nid_size; /* size of each NID, by bytes */
2576 __u8 mne_nid_count; /* # of NIDs in buffer */
2578 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2583 CONFIG_T_CONFIG = 0,
2584 CONFIG_T_SPTLRPC = 1,
2585 CONFIG_T_RECOVER = 2,
2586 CONFIG_T_PARAMS = 3,
2587 CONFIG_T_NODEMAP = 4,
2588 CONFIG_T_BARRIER = 5,
2592 struct mgs_config_body {
2593 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2594 __u64 mcb_offset; /* next index of config log to request */
2595 __u16 mcb_type; /* type of log: CONFIG_T_[CONFIG|RECOVER] */
2596 __u8 mcb_nm_cur_pass;
2597 __u8 mcb_bits; /* bits unit size of config log */
2598 __u32 mcb_units; /* # of units for bulk transfer */
2601 struct mgs_config_res {
2602 __u64 mcr_offset; /* index of last config log */
2604 __u64 mcr_size; /* size of the log */
2605 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2609 /* Config marker flags (in config log) */
2610 #define CM_START 0x01
2612 #define CM_SKIP 0x04
2613 #define CM_UPGRADE146 0x08
2614 #define CM_EXCLUDE 0x10
2615 #define CM_START_SKIP (CM_START | CM_SKIP)
2618 __u32 cm_step; /* aka config version */
2620 __u32 cm_vers; /* lustre release version number */
2621 __u32 cm_padding; /* 64 bit align */
2622 __s64 cm_createtime; /*when this record was first created */
2623 __s64 cm_canceltime; /*when this record is no longer valid*/
2624 char cm_tgtname[MTI_NAME_MAXLEN];
2625 char cm_comment[MTI_NAME_MAXLEN];
2629 * Opcodes for multiple servers.
2633 /* OBD_LOG_CANCEL = 401, obsolete since 1.5 */
2634 /* OBD_QC_CALLBACK = 402, obsolete since 2.4 */
2637 OBD_FIRST_OPC = OBD_PING
2641 * llog contexts indices.
2643 * There is compatibility problem with indexes below, they are not
2644 * continuous and must keep their numbers for compatibility needs.
2645 * See LU-5218 for details.
2648 LLOG_CONFIG_ORIG_CTXT = 0,
2649 LLOG_CONFIG_REPL_CTXT = 1,
2650 LLOG_MDS_OST_ORIG_CTXT = 2,
2651 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2652 LLOG_SIZE_ORIG_CTXT = 4,
2653 LLOG_SIZE_REPL_CTXT = 5,
2654 LLOG_TEST_ORIG_CTXT = 8,
2655 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2656 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2657 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2658 /* for multiple changelog consumers */
2659 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2660 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2661 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log */
2662 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log */
2666 /** Identifier for a single log object */
2668 struct ost_id lgl_oi;
2670 } __attribute__((packed));
2672 /** Records written to the CATALOGS list */
2673 #define CATLIST "CATALOGS"
2675 struct llog_logid lci_logid;
2679 } __attribute__((packed));
2681 /* Log data record types - there is no specific reason that these need to
2682 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2684 #define LLOG_OP_MAGIC 0x10600000
2685 #define LLOG_OP_MASK 0xfff00000
2688 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2689 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2690 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2691 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2692 REINT_UNLINK, /* obsolete after 2.5.0 */
2693 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2695 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2696 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2698 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2699 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2700 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2701 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2702 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2703 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2704 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2705 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000,
2706 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2707 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2710 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2711 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2713 /** Log record header - stored in little endian order.
2714 * Each record must start with this struct, end with a llog_rec_tail,
2715 * and be a multiple of 256 bits in size.
2717 struct llog_rec_hdr {
2724 struct llog_rec_tail {
2729 /* Where data follow just after header */
2730 #define REC_DATA(ptr) \
2731 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2733 #define REC_DATA_LEN(rec) \
2734 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2735 sizeof(struct llog_rec_tail))
2737 struct llog_logid_rec {
2738 struct llog_rec_hdr lid_hdr;
2739 struct llog_logid lid_id;
2743 struct llog_rec_tail lid_tail;
2744 } __attribute__((packed));
2746 struct llog_unlink_rec {
2747 struct llog_rec_hdr lur_hdr;
2751 struct llog_rec_tail lur_tail;
2752 } __attribute__((packed));
2754 struct llog_unlink64_rec {
2755 struct llog_rec_hdr lur_hdr;
2756 struct lu_fid lur_fid;
2757 __u32 lur_count; /* to destroy the lost precreated */
2761 struct llog_rec_tail lur_tail;
2762 } __attribute__((packed));
2764 struct llog_setattr64_rec {
2765 struct llog_rec_hdr lsr_hdr;
2766 struct ost_id lsr_oi;
2772 struct llog_rec_tail lsr_tail;
2773 } __attribute__((packed));
2775 /* Extended to support project quota */
2776 struct llog_setattr64_rec_v2 {
2777 struct llog_rec_hdr lsr_hdr;
2778 struct ost_id lsr_oi;
2785 __u32 lsr_layout_version;
2788 struct llog_rec_tail lsr_tail;
2789 } __attribute__((packed));
2791 struct llog_size_change_rec {
2792 struct llog_rec_hdr lsc_hdr;
2793 struct ll_fid lsc_fid;
2798 struct llog_rec_tail lsc_tail;
2799 } __attribute__((packed));
2801 #define CHANGELOG_MAGIC 0xca103000
2803 /** \a changelog_rec_type's that can't be masked */
2804 #define CHANGELOG_MINMASK (1 << CL_MARK)
2805 /** bits covering all \a changelog_rec_type's */
2806 #define CHANGELOG_ALLMASK 0XFFFFFFFF
2807 /** default \a changelog_rec_type mask. Allow all of them, except
2808 * CL_ATIME since it can really be time consuming, and not necessary
2810 * Remove also CL_OPEN, CL_GETXATTR and CL_DN_OPEN from default list as it can
2811 * be costly and only necessary for audit purpose.
2813 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & \
2814 ~(1 << CL_ATIME | 1 << CL_OPEN | 1 << CL_GETXATTR | \
2817 /* changelog llog name, needed by client replicators */
2818 #define CHANGELOG_CATALOG "changelog_catalog"
2820 struct changelog_setinfo {
2823 } __attribute__((packed));
2825 /** changelog record */
2826 struct llog_changelog_rec {
2827 struct llog_rec_hdr cr_hdr;
2828 struct changelog_rec cr; /**< Variable length field */
2829 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2830 } __attribute__((packed));
2832 #define CHANGELOG_USER_PREFIX "cl"
2834 struct llog_changelog_user_rec {
2835 struct llog_rec_hdr cur_hdr;
2837 /* only intended to be used in relative time comparisons to
2838 * detect idle users */
2841 struct llog_rec_tail cur_tail;
2842 } __attribute__((packed));
2844 enum agent_req_status {
2852 static inline const char *agent_req_status2name(enum agent_req_status ars)
2870 struct llog_agent_req_rec {
2871 struct llog_rec_hdr arr_hdr; /**< record header */
2872 __u32 arr_status; /**< status of the request */
2874 * agent_req_status */
2875 __u32 arr_archive_id; /**< backend archive number */
2876 __u64 arr_flags; /**< req flags */
2877 __u64 arr_compound_id; /**< compound cookie, ignored */
2878 __u64 arr_req_create; /**< req. creation time */
2879 __u64 arr_req_change; /**< req. status change time */
2880 struct hsm_action_item arr_hai; /**< req. to the agent */
2881 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2882 } __attribute__((packed));
2884 /* Old llog gen for compatibility */
2888 } __attribute__((packed));
2890 struct llog_gen_rec {
2891 struct llog_rec_hdr lgr_hdr;
2892 struct llog_gen lgr_gen;
2896 struct llog_rec_tail lgr_tail;
2899 /* flags for the logs */
2901 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2902 LLOG_F_IS_CAT = 0x2,
2903 LLOG_F_IS_PLAIN = 0x4,
2904 LLOG_F_EXT_JOBID = 0x8,
2905 LLOG_F_IS_FIXSIZE = 0x10,
2906 LLOG_F_EXT_EXTRA_FLAGS = 0x20,
2907 LLOG_F_EXT_X_UIDGID = 0x40,
2908 LLOG_F_EXT_X_NID = 0x80,
2909 LLOG_F_EXT_X_OMODE = 0x100,
2910 LLOG_F_EXT_X_XATTR = 0x200,
2912 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
2913 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
2914 * because the catlog record is usually fixed size, but its plain
2915 * log record can be variable */
2916 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID | LLOG_F_EXT_EXTRA_FLAGS |
2917 LLOG_F_EXT_X_UIDGID | LLOG_F_EXT_X_NID |
2918 LLOG_F_EXT_X_OMODE | LLOG_F_EXT_X_XATTR,
2921 /* means first record of catalog */
2923 LLOG_CAT_FIRST = -1,
2926 /* On-disk header structure of each log object, stored in little endian order */
2927 #define LLOG_MIN_CHUNK_SIZE 8192
2928 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
2929 * - sizeof(llh_bitmap) */
2930 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
2931 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2933 struct llog_log_hdr {
2934 struct llog_rec_hdr llh_hdr;
2935 __s64 llh_timestamp;
2937 __u32 llh_bitmap_offset;
2940 /* for a catalog the first/oldest and still in-use plain slot is just
2941 * next to it. It will serve as the upper limit after Catalog has
2944 struct obd_uuid llh_tgtuuid;
2945 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
2946 /* These fields must always be at the end of the llog_log_hdr.
2947 * Note: llh_bitmap size is variable because llog chunk size could be
2948 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
2949 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
2950 * llh_tail should only be refered by LLOG_HDR_TAIL().
2951 * But this structure is also used by client/server llog interface
2952 * (see llog_client.c), it will be kept in its original way to avoid
2953 * compatiblity issue. */
2954 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
2955 struct llog_rec_tail llh_tail;
2956 } __attribute__((packed));
2957 #undef LLOG_HEADER_SIZE
2958 #undef LLOG_BITMAP_BYTES
2960 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2961 llh->llh_bitmap_offset - \
2962 sizeof(llh->llh_tail)) * 8)
2963 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
2964 (llh)->llh_bitmap_offset)
2965 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
2966 llh->llh_hdr.lrh_len - \
2967 sizeof(llh->llh_tail)))
2969 /** log cookies are used to reference a specific log file and a record therein,
2970 and pass record offset from llog_process_thread to llog_write */
2971 struct llog_cookie {
2973 struct llog_logid lgc_lgl;
2979 } __attribute__((packed));
2981 /** llog protocol */
2982 enum llogd_rpc_ops {
2983 LLOG_ORIGIN_HANDLE_CREATE = 501,
2984 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
2985 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
2986 /* LLOG_ORIGIN_HANDLE_WRITE_REC = 504, Obsolete by 2.1. */
2987 /* LLOG_ORIGIN_HANDLE_CLOSE = 505, Obsolete by 1.8. */
2988 /* LLOG_ORIGIN_CONNECT = 506, Obsolete by 2.4. */
2989 /* LLOG_CATINFO = 507, Obsolete by 2.3. */
2990 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
2991 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* Obsolete by 2.11. */
2993 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
2997 struct llog_logid lgd_logid;
2999 __u32 lgd_llh_flags;
3001 __u32 lgd_saved_index;
3003 __u64 lgd_cur_offset;
3004 } __attribute__((packed));
3006 struct llogd_conn_body {
3007 struct llog_gen lgdc_gen;
3008 struct llog_logid lgdc_logid;
3009 __u32 lgdc_ctxt_idx;
3010 } __attribute__((packed));
3012 /* Note: 64-bit types are 64-bit aligned in structure */
3014 __u64 o_valid; /* hot fields in this obdo */
3017 __u64 o_size; /* o_size-o_blocks == ost_lvb */
3021 __u64 o_blocks; /* brw: cli sent cached bytes */
3024 /* 32-bit fields start here: keep an even number of them via padding */
3025 __u32 o_blksize; /* optimal IO blocksize */
3026 __u32 o_mode; /* brw: cli sent cache remain */
3030 __u32 o_nlink; /* brw: checksum */
3032 __u32 o_misc; /* brw: o_dropped */
3034 __u64 o_ioepoch; /* epoch in ost writes */
3035 __u32 o_stripe_idx; /* holds stripe idx */
3037 struct lustre_handle o_handle; /* brw: lock handle to prolong
3039 /* Originally, the field is llog_cookie for destroy with unlink cookie
3040 * from MDS, it is obsolete in 2.8. Then reuse it by client to transfer
3041 * layout and PFL information in IO, setattr RPCs. Since llog_cookie is
3042 * not used on wire any longer, remove it from the obdo, then it can be
3043 * enlarged freely in the further without affect related RPCs.
3045 * sizeof(ost_layout) + sieof(__u32) == sizeof(llog_cookie). */
3046 struct ost_layout o_layout;
3047 __u32 o_layout_version;
3051 __u64 o_data_version; /* getattr: sum of iversion for
3053 * brw: grant space consumed on
3054 * the client for the write */
3056 __u32 o_padding_4; /* also fix
3057 * lustre_swab_obdo() */
3062 #define o_dirty o_blocks
3063 #define o_undirty o_mode
3064 #define o_dropped o_misc
3065 #define o_cksum o_nlink
3066 #define o_grant_used o_data_version
3068 struct lfsck_request {
3080 __u16 lr_async_windows;
3082 struct lu_fid lr_fid;
3083 struct lu_fid lr_fid2;
3091 struct lfsck_reply {
3098 LE_LASTID_REBUILDING = 1,
3099 LE_LASTID_REBUILT = 2,
3105 /* LE_FID_ACCESSED = 8, moved to lfsck_events_local */
3107 LE_CONDITIONAL_DESTROY = 10,
3108 LE_PAIRS_VERIFY = 11,
3109 LE_SET_LMV_MASTER = 15,
3110 LE_SET_LMV_SLAVE = 16,
3113 enum lfsck_event_flags {
3114 LEF_TO_OST = 0x00000001,
3115 LEF_FROM_OST = 0x00000002,
3116 LEF_SET_LMV_HASH = 0x00000004,
3117 LEF_SET_LMV_ALL = 0x00000008,
3118 LEF_RECHECK_NAME_HASH = 0x00000010,
3119 LEF_QUERY_ALL = 0x00000020,
3122 /* request structure for OST's */
3127 /* Key for FIEMAP to be used in get_info calls */
3128 struct ll_fiemap_info_key {
3130 struct obdo lfik_oa;
3131 struct fiemap lfik_fiemap;
3134 #define IDX_INFO_MAGIC 0x3D37CC37
3136 /* Index file transfer through the network. The server serializes the index into
3137 * a byte stream which is sent to the client via a bulk transfer */
3141 /* reply: see idx_info_flags below */
3144 /* request & reply: number of lu_idxpage (to be) transferred */
3148 /* request: requested attributes passed down to the iterator API */
3151 /* request & reply: index file identifier (FID) */
3152 struct lu_fid ii_fid;
3154 /* reply: version of the index file before starting to walk the index.
3155 * Please note that the version can be modified at any time during the
3159 /* request: hash to start with:
3160 * reply: hash of the first entry of the first lu_idxpage and hash
3161 * of the entry to read next if any */
3162 __u64 ii_hash_start;
3165 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3169 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3178 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3180 /* List of flags used in idx_info::ii_flags */
3181 enum idx_info_flags {
3182 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3183 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3184 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3185 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3186 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
3189 #define LIP_MAGIC 0x8A6D6B6C
3191 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3193 /* 16-byte header */
3196 __u16 lip_nr; /* number of entries in the container */
3197 __u64 lip_pad0; /* additional padding for future use */
3199 /* key/record pairs are stored in the remaining 4080 bytes.
3200 * depending upon the flags in idx_info::ii_flags, each key/record
3201 * pair might be preceded by:
3203 * - the key size (II_FL_VARKEY is set)
3204 * - the record size (II_FL_VARREC is set)
3206 * For the time being, we only support fixed-size key & record. */
3207 char lip_entries[0];
3210 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3212 /* Gather all possible type associated with a 4KB container */
3214 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3215 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3216 char lp_array[LU_PAGE_SIZE];
3219 /* security opcodes */
3222 SEC_CTX_INIT_CONT = 802,
3225 SEC_FIRST_OPC = SEC_CTX_INIT
3229 * capa related definitions
3231 #define CAPA_HMAC_MAX_LEN 64
3232 #define CAPA_HMAC_KEY_MAX_LEN 56
3234 /* NB take care when changing the sequence of elements this struct,
3235 * because the offset info is used in find_capa() */
3236 struct lustre_capa {
3237 struct lu_fid lc_fid; /** fid */
3238 __u64 lc_opc; /** operations allowed */
3239 __u64 lc_uid; /** file owner */
3240 __u64 lc_gid; /** file group */
3241 __u32 lc_flags; /** HMAC algorithm & flags */
3242 __u32 lc_keyid; /** key# used for the capability */
3243 __u32 lc_timeout; /** capa timeout value (sec) */
3244 __u32 lc_expiry; /** expiry time (sec) */
3245 __u8 lc_hmac[CAPA_HMAC_MAX_LEN]; /** HMAC */
3246 } __attribute__((packed));
3248 /** lustre_capa::lc_opc */
3250 CAPA_OPC_BODY_WRITE = 1<<0, /**< write object data */
3251 CAPA_OPC_BODY_READ = 1<<1, /**< read object data */
3252 CAPA_OPC_INDEX_LOOKUP = 1<<2, /**< lookup object fid */
3253 CAPA_OPC_INDEX_INSERT = 1<<3, /**< insert object fid */
3254 CAPA_OPC_INDEX_DELETE = 1<<4, /**< delete object fid */
3255 CAPA_OPC_OSS_WRITE = 1<<5, /**< write oss object data */
3256 CAPA_OPC_OSS_READ = 1<<6, /**< read oss object data */
3257 CAPA_OPC_OSS_TRUNC = 1<<7, /**< truncate oss object */
3258 CAPA_OPC_OSS_DESTROY = 1<<8, /**< destroy oss object */
3259 CAPA_OPC_META_WRITE = 1<<9, /**< write object meta data */
3260 CAPA_OPC_META_READ = 1<<10, /**< read object meta data */
3263 #define CAPA_OPC_OSS_RW (CAPA_OPC_OSS_READ | CAPA_OPC_OSS_WRITE)
3264 #define CAPA_OPC_MDS_ONLY \
3265 (CAPA_OPC_BODY_WRITE | CAPA_OPC_BODY_READ | CAPA_OPC_INDEX_LOOKUP | \
3266 CAPA_OPC_INDEX_INSERT | CAPA_OPC_INDEX_DELETE)
3267 #define CAPA_OPC_OSS_ONLY \
3268 (CAPA_OPC_OSS_WRITE | CAPA_OPC_OSS_READ | CAPA_OPC_OSS_TRUNC | \
3269 CAPA_OPC_OSS_DESTROY)
3270 #define CAPA_OPC_MDS_DEFAULT ~CAPA_OPC_OSS_ONLY
3271 #define CAPA_OPC_OSS_DEFAULT ~(CAPA_OPC_MDS_ONLY | CAPA_OPC_OSS_ONLY)
3273 /* lustre_capa::lc_hmac_alg */
3275 CAPA_HMAC_ALG_SHA1 = 1, /**< sha1 algorithm */
3279 #define CAPA_FL_MASK 0x00ffffff
3280 #define CAPA_HMAC_ALG_MASK 0xff000000
3282 struct lustre_capa_key {
3283 __u64 lk_seq; /**< mds# */
3284 __u32 lk_keyid; /**< key# */
3286 __u8 lk_key[CAPA_HMAC_KEY_MAX_LEN]; /**< key */
3287 } __attribute__((packed));
3289 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3290 #define LINK_EA_MAGIC 0x11EAF1DFUL
3291 struct link_ea_header {
3294 __u64 leh_len; /* total size */
3295 __u32 leh_overflow_time;
3299 /** Hardlink data is name and parent fid.
3300 * Stored in this crazy struct for maximum packing and endian-neutrality
3302 struct link_ea_entry {
3303 /** __u16 stored big-endian, unaligned */
3304 unsigned char lee_reclen[2];
3305 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3307 }__attribute__((packed));
3309 /** fid2path request/reply structure */
3310 struct getinfo_fid2path {
3311 struct lu_fid gf_fid;
3317 struct lu_fid gf_root_fid[0];
3319 } __attribute__((packed));
3321 /** path2parent request/reply structures */
3323 struct lu_fid gp_fid; /**< parent FID */
3324 __u32 gp_linkno; /**< hardlink number */
3325 __u32 gp_name_size; /**< size of the name field */
3326 char gp_name[0]; /**< zero-terminated link name */
3327 } __attribute__((packed));
3329 enum layout_intent_opc {
3330 LAYOUT_INTENT_ACCESS = 0, /** generic access */
3331 LAYOUT_INTENT_READ = 1, /** not used */
3332 LAYOUT_INTENT_WRITE = 2, /** write file, for comp layout */
3333 LAYOUT_INTENT_GLIMPSE = 3, /** not used */
3334 LAYOUT_INTENT_TRUNC = 4, /** truncate file, for comp layout */
3335 LAYOUT_INTENT_RELEASE = 5, /** reserved for HSM release */
3336 LAYOUT_INTENT_RESTORE = 6, /** reserved for HSM restore */
3339 /* enqueue layout lock with intent */
3340 struct layout_intent {
3341 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3343 struct lu_extent li_extent;
3344 } __attribute__((packed));
3347 * On the wire version of hsm_progress structure.
3349 * Contains the userspace hsm_progress and some internal fields.
3351 struct hsm_progress_kernel {
3352 /* Field taken from struct hsm_progress */
3353 struct lu_fid hpk_fid;
3355 struct hsm_extent hpk_extent;
3357 __u16 hpk_errval; /* positive val */
3359 /* Additional fields */
3360 __u64 hpk_data_version;
3362 } __attribute__((packed));
3365 * OUT_UPDATE RPC Format
3367 * During the cross-ref operation, the Master MDT, which the client send the
3368 * request to, will disassembly the operation into object updates, then OSP
3369 * will send these updates to the remote MDT to be executed.
3371 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3372 * operation and does a type of modification to an object.
3380 * update (ub_count-th)
3382 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3387 * rc [+ buffers] (1st)
3388 * rc [+ buffers] (2st)
3390 * rc [+ buffers] (nr_count-th)
3392 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3393 * be equal to ub_count.
3397 * Type of each update, if adding/deleting update, please also update
3398 * update_opcode in lustre/target/out_lib.c.
3410 OUT_INDEX_LOOKUP = 9,
3411 OUT_INDEX_INSERT = 10,
3412 OUT_INDEX_DELETE = 11,
3418 OUT_XATTR_LIST = 17,
3423 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3424 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3425 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3426 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3429 struct object_update_param {
3430 __u16 oup_len; /* length of this parameter */
3437 struct object_update {
3438 __u16 ou_type; /* enum update_type */
3439 __u16 ou_params_count; /* update parameters count */
3440 __u32 ou_result_size; /* how many bytes can return */
3441 __u32 ou_flags; /* enum update_flag */
3442 __u32 ou_padding1; /* padding 1 */
3443 __u64 ou_batchid; /* op transno on master */
3444 struct lu_fid ou_fid; /* object to be updated */
3445 struct object_update_param ou_params[0]; /* update params */
3448 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3449 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3450 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3451 /* Hold object_updates sending to the remote OUT in single RPC */
3452 struct object_update_request {
3454 __u16 ourq_count; /* number of ourq_updates[] */
3456 struct object_update ourq_updates[0];
3459 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3460 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3461 /* Header for updates request between MDTs */
3462 struct out_update_header {
3465 __u32 ouh_inline_length;
3466 __u32 ouh_reply_size;
3467 __u32 ouh_inline_data[0];
3470 struct out_update_buffer {
3475 /* the result of object update */
3476 struct object_update_result {
3483 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3484 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3485 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3486 /* Hold object_update_results being replied from the remote OUT. */
3487 struct object_update_reply {
3494 /* read update result */
3495 struct out_read_reply {
3502 /** layout swap request structure
3503 * fid1 and fid2 are in mdt_body
3505 struct mdc_swap_layouts {
3507 } __attribute__((packed));
3509 #define INLINE_RESYNC_ARRAY_SIZE 15
3510 struct close_data_resync_done {
3512 __u32 resync_ids_inline[INLINE_RESYNC_ARRAY_SIZE];
3516 struct lustre_handle cd_handle;
3517 struct lu_fid cd_fid;
3518 __u64 cd_data_version;
3520 __u64 cd_reserved[8];
3521 struct close_data_resync_done cd_resync;
3527 /* Update llog format */
3529 struct lu_fid uop_fid;
3531 __u16 uop_param_count;
3532 __u16 uop_params_off[0];
3536 struct update_op uops_op[0];
3539 struct update_params {
3540 struct object_update_param up_params[0];
3543 enum update_records_flag {
3544 UPDATE_RECORD_CONTINUE = 1 >> 0,
3547 * This is the update record format used to store the updates in
3548 * disk. All updates of the operation will be stored in ur_ops.
3549 * All of parameters for updates of the operation will be stored
3551 * To save the space of the record, parameters in ur_ops will only
3552 * remember their offset in ur_params, so to avoid storing duplicate
3553 * parameters in ur_params, which can help us save a lot space for
3554 * operation like creating striped directory.
3556 struct update_records {
3557 __u64 ur_master_transno;
3560 /* If the operation includes multiple updates, then ur_index
3561 * means the index of the update inside the whole updates. */
3563 __u32 ur_update_count;
3564 __u32 ur_param_count;
3565 struct update_ops ur_ops;
3566 /* Note ur_ops has a variable size, so comment out
3567 * the following ur_params, in case some use it directly
3568 * update_records->ur_params
3570 * struct update_params ur_params;
3574 struct llog_update_record {
3575 struct llog_rec_hdr lur_hdr;
3576 struct update_records lur_update_rec;
3577 /* Note ur_update_rec has a variable size, so comment out
3578 * the following ur_tail, in case someone use it directly
3580 * struct llog_rec_tail lur_tail;
3584 /* sepol string format is:
3585 * <1-digit for SELinux status>:<policy name>:<policy version>:<policy hash>
3587 /* Max length of the sepol string
3588 * Should be large enough to contain a sha512sum of the policy
3590 #define SELINUX_MODE_LEN 1
3591 #define SELINUX_POLICY_VER_LEN 3 /* 3 chars to leave room for the future */
3592 #define SELINUX_POLICY_HASH_LEN 64
3593 #define LUSTRE_NODEMAP_SEPOL_LENGTH (SELINUX_MODE_LEN + NAME_MAX + \
3594 SELINUX_POLICY_VER_LEN + \
3595 SELINUX_POLICY_HASH_LEN + 3)
3597 /* nodemap records, uses 32 byte record length */
3598 #define LUSTRE_NODEMAP_NAME_LENGTH 16
3599 struct nodemap_cluster_rec {
3600 char ncr_name[LUSTRE_NODEMAP_NAME_LENGTH + 1];
3604 __u32 ncr_squash_uid;
3605 __u32 ncr_squash_gid;
3608 /* lnet_nid_t is 8 bytes */
3609 struct nodemap_range_rec {
3610 lnet_nid_t nrr_start_nid;
3611 lnet_nid_t nrr_end_nid;
3616 struct nodemap_id_rec {
3624 struct nodemap_global_rec {
3635 struct nodemap_cluster_rec ncr;
3636 struct nodemap_range_rec nrr;
3637 struct nodemap_id_rec nir;
3638 struct nodemap_global_rec ngr;
3641 /* This is the lu_ladvise struct which goes out on the wire.
3642 * Corresponds to the userspace arg llapi_lu_ladvise.
3643 * value[1-4] are unspecified fields, used differently by different advices */
3645 __u16 lla_advice; /* advice type */
3646 __u16 lla_value1; /* values for different advice types */
3648 __u64 lla_start; /* first byte of extent for advice */
3649 __u64 lla_end; /* last byte of extent for advice */
3654 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3655 * arg llapi_ladvise_hdr.
3656 * value[1-3] are unspecified fields, used differently by different advices */
3657 struct ladvise_hdr {
3658 __u32 lah_magic; /* LADVISE_MAGIC */
3659 __u32 lah_count; /* number of advices */
3660 __u64 lah_flags; /* from enum ladvise_flag */
3661 __u32 lah_value1; /* unused */
3662 __u32 lah_value2; /* unused */
3663 __u64 lah_value3; /* unused */
3664 struct lu_ladvise lah_advise[0]; /* advices in this header */
3667 #if defined(__cplusplus)