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/
31 * Lustre wire protocol definitions.
34 /** \defgroup lustreidl lustreidl
36 * Lustre wire protocol definitions.
38 * ALL structs passing over the wire should be declared here. Structs
39 * that are used in interfaces with userspace should go in lustre_user.h.
41 * All structs being declared here should be built from simple fixed-size
42 * types defined in linux/types.h or be built from other types or
43 * structs also declared in this file. Similarly, all flags and magic
44 * values in those structs should also be declared here. This ensures
45 * that the Lustre wire protocol is not influenced by external dependencies.
47 * The only other acceptable items in this file are VERY SIMPLE accessor
48 * functions to avoid callers grubbing inside the structures. Nothing that
49 * depends on external functions or definitions should be in here.
51 * Structs must be properly aligned to put 64-bit values on an 8-byte
52 * boundary. Any structs being added here must also be added to
53 * utils/wirecheck.c and "make newwiretest" run to regenerate the
54 * utils/wiretest.c sources. This allows us to verify that wire structs
55 * have the proper alignment/size on all architectures.
57 * DO NOT CHANGE any of the structs, flags, values declared here and used
58 * in released Lustre versions. Some structs may have padding fields that
59 * can be used. Some structs might allow addition at the end (verify this
60 * in the code to ensure that new/old clients that see this larger struct
61 * do not fail, otherwise you need to implement protocol compatibility).
66 #ifndef _LUSTRE_IDL_H_
67 #define _LUSTRE_IDL_H_
69 #include <asm/byteorder.h>
70 #include <linux/errno.h>
71 #include <linux/fiemap.h>
72 #include <linux/types.h>
73 #include <linux/lnet/lnet-types.h>
74 #include <linux/lustre/lustre_user.h>
75 #include <linux/lustre/lustre_ver.h>
77 #if defined(__cplusplus)
84 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
85 * FOO_REPLY_PORTAL is for incoming replies on the FOO
86 * FOO_BULK_PORTAL is for incoming bulk on the FOO
89 #define CONNMGR_REQUEST_PORTAL 1
90 #define CONNMGR_REPLY_PORTAL 2
91 /* #define OSC_REQUEST_PORTAL 3*/
92 #define OSC_REPLY_PORTAL 4
93 /*#define OSC_BULK_PORTAL 5*/
94 #define OST_IO_PORTAL 6
95 #define OST_CREATE_PORTAL 7
96 #define OST_BULK_PORTAL 8
97 /*#define MDC_REQUEST_PORTAL 9*/
98 #define MDC_REPLY_PORTAL 10
99 /*#define MDC_BULK_PORTAL 11*/
100 #define MDS_REQUEST_PORTAL 12
101 #define MDS_IO_PORTAL 13
102 #define MDS_BULK_PORTAL 14
103 #define LDLM_CB_REQUEST_PORTAL 15
104 #define LDLM_CB_REPLY_PORTAL 16
105 #define LDLM_CANCEL_REQUEST_PORTAL 17
106 #define LDLM_CANCEL_REPLY_PORTAL 18
107 /*#define PTLBD_REQUEST_PORTAL 19*/
108 /*#define PTLBD_REPLY_PORTAL 20*/
109 /*#define PTLBD_BULK_PORTAL 21*/
110 #define MDS_SETATTR_PORTAL 22
111 #define MDS_READPAGE_PORTAL 23
112 #define OUT_PORTAL 24
113 #define MGC_REPLY_PORTAL 25
114 #define MGS_REQUEST_PORTAL 26
115 #define MGS_REPLY_PORTAL 27
116 #define OST_REQUEST_PORTAL 28
117 #define FLD_REQUEST_PORTAL 29
118 #define SEQ_METADATA_PORTAL 30
119 #define SEQ_DATA_PORTAL 31
120 #define SEQ_CONTROLLER_PORTAL 32
121 #define MGS_BULK_PORTAL 33
122 /* #define DVS_PORTAL 63 */
123 /* reserved for Cray DVS - spitzcor@cray.com, roe@cray.com, n8851@cray.com */
126 * Describes a range of sequence, lsr_start is included but lsr_end is
128 * Same structure is used in fld module where lsr_index field holds mdt id
131 struct lu_seq_range {
138 struct lu_seq_range_array {
141 struct lu_seq_range lsra_lsr[0];
144 #define LU_SEQ_RANGE_MDT 0x0
145 #define LU_SEQ_RANGE_OST 0x1
146 #define LU_SEQ_RANGE_ANY 0x3
148 #define LU_SEQ_RANGE_MASK 0x3
150 /** \defgroup lu_fid lu_fid
153 extern void lustre_lma_swab(struct lustre_mdt_attrs *lma);
154 extern void lustre_lma_init(struct lustre_mdt_attrs *lma,
155 const struct lu_fid *fid,
156 __u32 compat, __u32 incompat);
157 extern void lustre_loa_swab(struct lustre_ost_attrs *loa,
159 extern void lustre_loa_init(struct lustre_ost_attrs *loa,
160 const struct lu_fid *fid,
161 __u32 compat, __u32 incompat);
163 /* copytool can use any nonnegative integer to represent archive-Ids during
164 * register with MDT thru kuc.
165 * archive num = 0 => all
166 * archive num from 1 to MAX_U32
168 #define LL_HSM_ORIGIN_MAX_ARCHIVE (sizeof(__u32) * 8)
169 /* the max count of archive ids that one agent can support */
170 #define LL_HSM_MAX_ARCHIVES_PER_AGENT 1024
173 * HSM on-disk attributes stored in a separate xattr.
176 /** Bitfield for supported data in this structure. For future use. */
179 /** HSM flags, see hsm_flags enum below */
181 /** backend archive id associated with the file */
183 /** version associated with the last archiving, if any */
186 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
192 /** LASTID file has zero OID */
193 LUSTRE_FID_LASTID_OID = 0UL,
194 /** initial fid id value */
195 LUSTRE_FID_INIT_OID = 1UL
199 * Different FID Format
200 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
203 * File IDentifier generated by client from range allocated by the seq service.
204 * First 0x400 sequences [2^33, 2^33 + 0x400] are reserved for system use. Note
205 * that on ldiskfs MDTs that IGIF FIDs can use inode numbers starting at 12,
206 * but this is in the IGIF SEQ rangeand does not conflict with assigned FIDs.
209 * Inode and Generation In FID, a surrogate FID used to globally identify an
210 * existing object on OLD formatted MDT file system. This would only be used on
211 * MDT0 in a DNE filesystem, because there are not expected to be any OLD
212 * formatted DNE filesystems. Belongs to a sequence in [12, 2^32 - 1] range,
213 * where sequence number is inode number, and inode generation is used as OID.
214 * NOTE: This assumes no more than 2^32-1 inodes exist in the MDT filesystem,
215 * which is the maximum possible for an ldiskfs backend. NOTE: This assumes
216 * that the reserved ext3/ext4/ldiskfs inode numbers [0-11] are never visible
217 * to clients, which has always been true.
220 * Object ID in FID, a surrogate FID used to globally identify an existing
221 * object on OLD formatted OST file system. Belongs to a sequence in
222 * [2^32, 2^33 - 1]. Sequence number is calculated as:
223 * 1 << 32 | (ost_index << 16) | ((objid >> 32) & 0xffff)
224 * that is, SEQ consists of 16-bit OST index, and higher 16 bits of object ID.
225 * The generation of unique SEQ values per OST allows the IDIF FIDs to be
226 * identified in the FLD correctly. The OID field is calculated as:
228 * that is, it consists of lower 32 bits of object ID. NOTE This assumes that
229 * no more than 2^48-1 objects have ever been created on an OST, and that no
230 * more than 65535 OSTs are in use. Both are very reasonable assumptions (can
231 * uniquely map all objects on an OST that created 1M objects per second for 9
232 * years, or combinations thereof).
235 * Surrogate FID used to identify an existing object on OLD formatted OST
236 * filesystem. Belongs to the reserved sequence 0, and is used internally prior
237 * to the introduction of FID-on-OST, at which point IDIF will be used to
238 * identify objects as residing on a specific OST.
241 * For Lustre Log objects the object sequence 1 is used. This is compatible with
242 * both OLD and NEW.1 namespaces, as this SEQ number is in the ext3/ldiskfs
243 * reserved inode range and does not conflict with IGIF sequence numbers.
246 * For testing OST IO performance the object sequence 2 is used. This is
247 * compatible with both OLD and NEW.1 namespaces, as this SEQ number is in the
248 * ext3/ldiskfs reserved inode range and does not conflict with IGIF sequence
251 * OST_MDT1 .. OST_MAX:
252 * For testing with multiple MDTs the object sequence 3 through 9 is used,
253 * allowing direct mapping of MDTs 1 through 7 respectively, for a total of 8
254 * MDTs including OST_MDT0. This matches the legacy CMD project "group"
255 * mappings. However, this SEQ range is only for testing prior to any production
256 * DNE release, as the objects in this range conflict across all OSTs, as the
257 * OST index is not part of the FID.
260 * For compatibility with existing OLD OST network protocol structures, the FID
261 * must map onto the o_id and o_gr in a manner that ensures existing objects are
262 * identified consistently for IO, as well as onto the lock namespace to ensure
263 * both IDIFs map onto the same objects for IO as well as resources in the DLM.
266 * resource[] = {o_id, o_seq, 0, 0}; // o_seq == 0 for production releases
268 * DLM NEW.1 FID (this is the same for both the MDT and OST):
269 * resource[] = {SEQ, OID, VER, HASH};
271 * Note that for mapping IDIF values to DLM resource names the o_id may be
272 * larger than the 2^33 reserved sequence numbers for IDIF, so it is possible
273 * for the o_id numbers to overlap FID SEQ numbers in the resource. However, in
274 * all production releases the OLD o_seq field is always zero, and all valid FID
275 * OID values are non-zero, so the lock resources will not collide.
277 * For objects within the IDIF range, group extraction (non-CMD) will be:
278 * o_id = (fid->f_seq & 0x7fff) << 16 | fid->f_oid;
279 * o_seq = 0; // formerly group number
283 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
284 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
285 * used for other purposes and not risk collisions with existing inodes.
288 FID_SEQ_OST_MDT0 = 0,
289 FID_SEQ_LLOG = 1, /* unnamed llogs */
291 FID_SEQ_UNUSED_START = 3,
292 FID_SEQ_UNUSED_END = 9,
293 FID_SEQ_LLOG_NAME = 10, /* named llogs */
296 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
297 FID_SEQ_IDIF = 0x100000000ULL,
298 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
299 /* Normal FID sequence starts from this value, i.e. 1<<33 */
300 FID_SEQ_START = 0x200000000ULL,
301 /* sequence for local pre-defined FIDs listed in local_oid */
302 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
303 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
304 /* sequence is used for local named objects FIDs generated
305 * by local_object_storage library */
306 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
307 /* Because current FLD will only cache the fid sequence, instead
308 * of oid on the client side, if the FID needs to be exposed to
309 * clients sides, it needs to make sure all of fids under one
310 * sequence will be located in one MDT. */
311 FID_SEQ_SPECIAL = 0x200000004ULL,
312 FID_SEQ_QUOTA = 0x200000005ULL,
313 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
314 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
315 FID_SEQ_LAYOUT_RBTREE = 0x200000008ULL,
316 /* sequence is used for update logs of cross-MDT operation */
317 FID_SEQ_UPDATE_LOG = 0x200000009ULL,
318 /* Sequence is used for the directory under which update logs
320 FID_SEQ_UPDATE_LOG_DIR = 0x20000000aULL,
321 FID_SEQ_NORMAL = 0x200000400ULL,
322 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
325 #define OBIF_OID_MAX_BITS 32
326 #define OBIF_MAX_OID (1ULL << OBIF_OID_MAX_BITS)
327 #define OBIF_OID_MASK ((1ULL << OBIF_OID_MAX_BITS) - 1)
328 #define IDIF_OID_MAX_BITS 48
329 #define IDIF_MAX_OID (1ULL << IDIF_OID_MAX_BITS)
330 #define IDIF_OID_MASK ((1ULL << IDIF_OID_MAX_BITS) - 1)
332 /** OID for FID_SEQ_SPECIAL */
334 /* Big Filesystem Lock to serialize rename operations */
335 FID_OID_SPECIAL_BFL = 1UL,
338 /** OID for FID_SEQ_DOT_LUSTRE */
339 enum dot_lustre_oid {
340 FID_OID_DOT_LUSTRE = 1UL,
341 FID_OID_DOT_LUSTRE_OBF = 2UL,
342 FID_OID_DOT_LUSTRE_LPF = 3UL,
345 /** OID for FID_SEQ_ROOT */
348 FID_OID_ECHO_ROOT = 2UL,
351 struct lu_orphan_rec {
352 /* The MDT-object's FID referenced by the orphan OST-object */
353 struct lu_fid lor_fid;
358 struct lu_orphan_ent {
359 /* The orphan OST-object's FID */
360 struct lu_fid loe_key;
361 struct lu_orphan_rec loe_rec;
364 struct lu_orphan_rec_v2 {
365 struct lu_orphan_rec lor_rec;
366 struct ost_layout lor_layout;
370 struct lu_orphan_ent_v2 {
371 /* The orphan OST-object's FID */
372 struct lu_fid loe_key;
373 struct lu_orphan_rec_v2 loe_rec;
376 struct lu_orphan_rec_v3 {
377 struct lu_orphan_rec lor_rec;
378 struct ost_layout lor_layout;
379 /* The OST-object declared layout version in PFID EA.*/
380 __u32 lor_layout_version;
381 /* The OST-object declared layout range (of version) in PFID EA.*/
387 struct lu_orphan_ent_v3 {
388 /* The orphan OST-object's FID */
389 struct lu_fid loe_key;
390 struct lu_orphan_rec_v3 loe_rec;
395 /** \defgroup lu_dir lu_dir
399 * Enumeration of possible directory entry attributes.
401 * Attributes follow directory entry header in the order they appear in this
404 enum lu_dirent_attrs {
407 LUDA_64BITHASH = 0x0004,
409 /* The following attrs are used for MDT internal only,
410 * not visible to client */
412 /* Something in the record is unknown, to be verified in further. */
413 LUDA_UNKNOWN = 0x0400,
414 /* Ignore this record, go to next directly. */
415 LUDA_IGNORE = 0x0800,
416 /* The system is upgraded, has beed or to be repaired (dryrun). */
417 LUDA_UPGRADE = 0x1000,
418 /* The dirent has been repaired, or to be repaired (dryrun). */
419 LUDA_REPAIR = 0x2000,
420 /* Only check but not repair the dirent inconsistency */
421 LUDA_VERIFY_DRYRUN = 0x4000,
422 /* Verify the dirent consistency */
423 LUDA_VERIFY = 0x8000,
426 #define LU_DIRENT_ATTRS_MASK 0xff00
429 * Layout of readdir pages, as transmitted on wire.
432 /** valid if LUDA_FID is set. */
433 struct lu_fid lde_fid;
434 /** a unique entry identifier: a hash or an offset. */
436 /** total record length, including all attributes. */
440 /** optional variable size attributes following this entry.
441 * taken from enum lu_dirent_attrs.
444 /** name is followed by the attributes indicated in ->ldp_attrs, in
445 * their natural order. After the last attribute, padding bytes are
446 * added to make ->lde_reclen a multiple of 8.
452 * Definitions of optional directory entry attributes formats.
454 * Individual attributes do not have their length encoded in a generic way. It
455 * is assumed that consumer of an attribute knows its format. This means that
456 * it is impossible to skip over an unknown attribute, except by skipping over all
457 * remaining attributes (by using ->lde_reclen), which is not too
458 * constraining, because new server versions will append new attributes at
459 * the end of an entry.
463 * Fid directory attribute: a fid of an object referenced by the entry. This
464 * will be almost always requested by the client and supplied by the server.
466 * Aligned to 8 bytes.
468 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
473 * Aligned to 2 bytes.
480 __u64 ldp_hash_start;
484 struct lu_dirent ldp_entries[0];
487 enum lu_dirpage_flags {
489 * dirpage contains no entry.
493 * last entry's lde_hash equals ldp_hash_end.
498 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
500 if (__le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
503 return dp->ldp_entries;
506 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
508 struct lu_dirent *next;
510 if (__le16_to_cpu(ent->lde_reclen) != 0)
511 next = ((void *)ent) + __le16_to_cpu(ent->lde_reclen);
518 static inline __kernel_size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
520 __kernel_size_t size;
522 if (attr & LUDA_TYPE) {
523 const __kernel_size_t align = sizeof(struct luda_type) - 1;
525 size = (sizeof(struct lu_dirent) + namelen + 1 + align) &
527 size += sizeof(struct luda_type);
529 size = sizeof(struct lu_dirent) + namelen + 1;
532 return (size + 7) & ~7;
535 static inline __u16 lu_dirent_type_get(struct lu_dirent *ent)
538 struct luda_type *lt;
541 if (__le32_to_cpu(ent->lde_attrs) & LUDA_TYPE) {
542 const unsigned int align = sizeof(struct luda_type) - 1;
544 len = __le16_to_cpu(ent->lde_namelen);
545 len = (len + align) & ~align;
546 lt = (void *)ent->lde_name + len;
547 type = __le16_to_cpu(lt->lt_type);
553 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
556 * MDS_READPAGE page size
558 * This is the directory page size packed in MDS_READPAGE RPC.
559 * It's different than PAGE_SIZE because the client needs to
560 * access the struct lu_dirpage header packed at the beginning of
561 * the "page" and without this there isn't any way to know find the
562 * lu_dirpage header is if client and server PAGE_SIZE differ.
564 #define LU_PAGE_SHIFT 12
565 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
566 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
568 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
572 struct lustre_handle {
575 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
577 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
579 return lh->cookie != 0;
582 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
583 const struct lustre_handle *lh2)
585 return lh1->cookie == lh2->cookie;
588 static inline void lustre_handle_copy(struct lustre_handle *tgt,
589 const struct lustre_handle *src)
591 tgt->cookie = src->cookie;
594 /* lustre_msg struct magic. DON'T use swabbed values of MAGIC as magic! */
595 enum lustre_msg_magic {
596 LUSTRE_MSG_MAGIC_V2 = 0x0BD00BD3,
597 LUSTRE_MSG_MAGIC_V2_SWABBED = 0xD30BD00B,
598 LUSTRE_MSG_MAGIC = LUSTRE_MSG_MAGIC_V2
601 /* flags for lm_flags */
603 MSGHDR_AT_SUPPORT = 0x1, /* adaptive timeouts, lm_cksum valid
604 * in early reply messages */
605 MSGHDR_CKSUM_INCOMPAT18 = 0x2, /* compat for 1.8, needs to be set well
606 * beyond 2.8.0 for compatibility */
609 #define lustre_msg lustre_msg_v2
610 /* we depend on this structure to be 8-byte aligned */
611 /* this type is only endian-adjusted in lustre_unpack_msg() */
612 struct lustre_msg_v2 {
613 __u32 lm_bufcount; /* number of buffers in lm_buflens[] */
614 __u32 lm_secflvr; /* 0 = no crypto, or sptlrpc security flavour */
615 __u32 lm_magic; /* RPC version magic = LUSTRE_MSG_MAGIC_V2 */
616 __u32 lm_repsize; /* size of preallocated reply buffer */
617 __u32 lm_cksum; /* CRC32 of ptlrpc_body early reply messages */
618 __u32 lm_flags; /* enum lustre_msghdr MSGHDR_* flags */
619 __u32 lm_padding_2; /* unused */
620 __u32 lm_padding_3; /* unused */
621 __u32 lm_buflens[0]; /* length of additional buffers in bytes,
622 * padded to a multiple of 8 bytes. */
624 * message buffers are packed after padded lm_buflens[] array,
625 * padded to a multiple of 8 bytes each to align contents.
629 /* ptlrpc_body packet pb_types */
630 #define PTL_RPC_MSG_REQUEST 4711 /* normal RPC request message */
631 #define PTL_RPC_MSG_ERR 4712 /* error reply if request unprocessed */
632 #define PTL_RPC_MSG_REPLY 4713 /* normal RPC reply message */
634 /* ptlrpc_body pb_version ((target_version << 16) | rpc_version) */
635 enum lustre_msg_version {
636 PTLRPC_MSG_VERSION = 0x00000003,
637 LUSTRE_VERSION_MASK = 0xffff0000,
638 LUSTRE_OBD_VERSION = 0x00010000,
639 LUSTRE_MDS_VERSION = 0x00020000,
640 LUSTRE_OST_VERSION = 0x00030000,
641 LUSTRE_DLM_VERSION = 0x00040000,
642 LUSTRE_LOG_VERSION = 0x00050000,
643 LUSTRE_MGS_VERSION = 0x00060000,
646 /* pb_flags that apply to all request messages */
647 /* #define MSG_LAST_REPLAY 0x0001 obsolete 2.0 => {REQ,LOCK}_REPLAY_DONE */
648 #define MSG_RESENT 0x0002 /* was previously sent, no reply seen */
649 #define MSG_REPLAY 0x0004 /* was processed, got reply, recovery */
650 /* #define MSG_AT_SUPPORT 0x0008 obsolete since 1.5, AT always enabled */
651 /* #define MSG_DELAY_REPLAY 0x0010 obsolete since 2.0 */
652 /* #define MSG_VERSION_REPLAY 0x0020 obsolete since 1.8.2, VBR always on */
653 #define MSG_REQ_REPLAY_DONE 0x0040 /* request replay over, locks next */
654 #define MSG_LOCK_REPLAY_DONE 0x0080 /* lock replay over, client done */
656 /* pb_op_flags for connect opcodes: MDS_CONNECT, OST_CONNECT, MGS_CONNECT */
657 #define MSG_CONNECT_RECOVERING 0x00000001 /* target is in recovery */
658 #define MSG_CONNECT_RECONNECT 0x00000002 /* tgt already has client import */
659 #define MSG_CONNECT_REPLAYABLE 0x00000004 /* target supports RPC replay */
660 /* #define MSG_CONNECT_PEER 0x00000008 obsolete since 1.2, removed in 1.5 */
661 #define MSG_CONNECT_LIBCLIENT 0x00000010 /* obsolete since 2.3, removed 2.6 */
662 #define MSG_CONNECT_INITIAL 0x00000020 /* first client connection attempt */
663 /* #define MSG_CONNECT_ASYNC 0x00000040 obsolete since 1.5 */
664 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
665 #define MSG_CONNECT_TRANSNO 0x00000100 /* client sent transno in replay */
667 /* number of previous object versions in pb_pre_versions[] */
668 #define PTLRPC_NUM_VERSIONS 4
669 /* without gss, ptlrpc_body is put at the first buffer. */
670 struct ptlrpc_body_v3 {
671 struct lustre_handle pb_handle;
672 __u32 pb_type; /* request/reply/err type: PTL_RPC_MSG_* */
673 __u32 pb_version; /* LUSTRE_*_VERSION | PTLRPC_MSG_VERSION */
674 __u32 pb_opc; /* RPC opcodes: MDS_*, OST_*, LDLM_, ... */
675 __u32 pb_status; /* negative Linux x86 error number */
676 __u64 pb_last_xid; /* highest replied XID w/o lower unreplied XID*/
677 __u16 pb_tag; /* multiple modifying RPCs virtual slot index */
680 __u64 pb_last_committed;/* rep: highest pb_transno committed to disk */
681 __u64 pb_transno; /* server-assigned transno for modifying RPCs */
682 __u32 pb_flags; /* req: MSG_* flags */
683 __u32 pb_op_flags; /* req: MSG_CONNECT_* flags */
684 __u32 pb_conn_cnt; /* connect instance of this client on server */
685 __u32 pb_timeout; /* req: max wait time; rep: service estimate */
686 __u32 pb_service_time; /* rep: server arrival to reply in seconds */
687 __u32 pb_limit; /* rep: dynamic DLM LRU lock count limit */
688 __u64 pb_slv; /* rep: dynamic DLM LRU server lock volume */
689 /* VBR: rep: previous pb_version(s) of objects modified by this RPC */
690 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
691 __u64 pb_mbits; /**< match bits for bulk request */
692 /* padding for future needs - fix lustre_swab_ptlrpc_body() also */
693 __u64 pb_padding64_0;
694 __u64 pb_padding64_1;
695 __u64 pb_padding64_2;
696 char pb_jobid[LUSTRE_JOBID_SIZE]; /* req: ASCII jobid from env + NUL */
698 #define ptlrpc_body ptlrpc_body_v3
700 struct ptlrpc_body_v2 {
701 struct lustre_handle pb_handle;
706 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
707 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
710 __u64 pb_last_committed;
715 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
716 __u32 pb_service_time; /* for rep, actual service time, also used for
717 net_latency of req */
720 /* VBR: pre-versions */
721 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
722 __u64 pb_mbits; /**< unused in V2 */
723 /* padding for future needs */
724 __u64 pb_padding64_0;
725 __u64 pb_padding64_1;
726 __u64 pb_padding64_2;
729 /* message body offset for lustre_msg_v2 */
730 /* ptlrpc body offset in all request/reply messages */
731 #define MSG_PTLRPC_BODY_OFF 0
733 /* normal request/reply message record offset */
734 #define REQ_REC_OFF 1
735 #define REPLY_REC_OFF 1
737 /* ldlm request message body offset */
738 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
739 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
741 /* ldlm intent lock message body offset */
742 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
743 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
745 /* ldlm reply message body offset */
746 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
747 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
749 /** only use in req->rq_{req,rep}_swab_mask */
750 #define MSG_PTLRPC_HEADER_OFF 31
753 #define OBD_CONNECT_RDONLY 0x1ULL /*client has read-only access*/
754 #define OBD_CONNECT_INDEX 0x2ULL /*connect specific LOV idx */
755 #define OBD_CONNECT_MDS 0x4ULL /*connect from MDT to OST */
756 #define OBD_CONNECT_GRANT 0x8ULL /*OSC gets grant at connect */
757 #define OBD_CONNECT_SRVLOCK 0x10ULL /*server takes locks for cli */
758 #define OBD_CONNECT_VERSION 0x20ULL /*Lustre versions in ocd */
759 #define OBD_CONNECT_REQPORTAL 0x40ULL /*Separate non-IO req portal */
760 #define OBD_CONNECT_ACL 0x80ULL /*access control lists */
761 #define OBD_CONNECT_XATTR 0x100ULL /*client use extended attr */
762 #define OBD_CONNECT_LARGE_ACL 0x200ULL /* more than 32 ACL entries */
763 #define OBD_CONNECT_TRUNCLOCK 0x400ULL /*locks on server for punch */
764 #define OBD_CONNECT_TRANSNO 0x800ULL /*replay sends init transno */
765 #define OBD_CONNECT_IBITS 0x1000ULL /* not checked in 2.11+ */
766 #define OBD_CONNECT_BARRIER 0x2000ULL /* write barrier */
767 #define OBD_CONNECT_ATTRFID 0x4000ULL /*Server can GetAttr By Fid*/
768 #define OBD_CONNECT_NODEVOH 0x8000ULL /*No open hndl on specl nodes*/
769 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Remote client, never used
770 * in production. Removed in
771 * 2.9. Keep this flag to
774 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Remote client by force,
775 * never used in production.
776 * Removed in 2.9. Keep this
777 * flag to avoid reusing.
779 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /*Max bytes per rpc */
780 #define OBD_CONNECT_QUOTA64 0x80000ULL /*Not used since 2.4 */
781 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /*MDS capability */
782 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /*OSS capability */
783 #define OBD_CONNECT_CANCELSET 0x400000ULL /*Early batched cancels. */
784 #define OBD_CONNECT_SOM 0x800000ULL /*Size on MDS */
785 #define OBD_CONNECT_AT 0x1000000ULL /*client uses AT */
786 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /*LRU resize feature. */
787 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /*MDS-MDS connection */
788 #define OBD_CONNECT_REAL 0x8000000ULL /* obsolete since 2.8 */
789 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /*Not used since 2.4 */
790 #define OBD_CONNECT_CKSUM 0x20000000ULL /*support several cksum algos*/
791 #define OBD_CONNECT_FID 0x40000000ULL /*FID is supported by server */
792 #define OBD_CONNECT_VBR 0x80000000ULL /*version based recovery */
793 #define OBD_CONNECT_LOV_V3 0x100000000ULL /*client supports LOV v3 EA */
794 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* support grant shrink */
795 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* don't reuse orphan objids */
796 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* preserved for large EA */
797 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
798 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client uses layout lock */
799 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* client supports 64-bits
801 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
802 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recovery support */
803 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* jobid in ptlrpc_body */
804 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create uses client umask */
805 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client handles -EINPROGRESS
806 * RPC error properly */
807 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL/* extra grant params used for
808 * finer space reservation */
809 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* for the fixed 1.8
810 * policy and 2.x server */
811 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable type of LVB */
812 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosecond timestamps */
813 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL/* lightweight connection */
814 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL/* short io */
815 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL/* pings not required */
816 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL/* improved flock deadlock detection */
817 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL/* create stripe disposition*/
818 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by fid won't pack
820 #define OBD_CONNECT_LFSCK 0x40000000000000ULL/* support online LFSCK */
821 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL/* close file in unlink */
822 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* support multiple modify
824 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
825 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
826 /* was OBD_CONNECT_LOCKAHEAD_OLD 0x1000000000000000ULL old lockahead 2.12-2.13*/
828 /** bulk matchbits is sent within ptlrpc_body */
829 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL
830 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
831 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
832 /* ocd_connect_flags2 flags */
833 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* set file security context at create */
834 #define OBD_CONNECT2_LOCKAHEAD 0x2ULL /* ladvise lockahead v2 */
835 #define OBD_CONNECT2_DIR_MIGRATE 0x4ULL /* migrate striped dir */
836 #define OBD_CONNECT2_SUM_STATFS 0x8ULL /* MDT return aggregated stats */
837 #define OBD_CONNECT2_OVERSTRIPING 0x10ULL /* OST overstriping support */
838 #define OBD_CONNECT2_FLR 0x20ULL /* FLR support */
839 #define OBD_CONNECT2_WBC_INTENTS 0x40ULL /* create/unlink/... intents for wbc, also operations under client-held parent locks */
840 #define OBD_CONNECT2_LOCK_CONVERT 0x80ULL /* IBITS lock convert support */
841 #define OBD_CONNECT2_ARCHIVE_ID_ARRAY 0x100ULL /* store HSM archive_id in array */
842 #define OBD_CONNECT2_INC_XID 0x200ULL /* Increasing xid */
843 #define OBD_CONNECT2_SELINUX_POLICY 0x400ULL /* has client SELinux policy */
844 #define OBD_CONNECT2_LSOM 0x800ULL /* LSOM support */
845 #define OBD_CONNECT2_PCC 0x1000ULL /* Persistent Client Cache */
846 #define OBD_CONNECT2_CRUSH 0x2000ULL /* crush hash striped directory */
847 #define OBD_CONNECT2_ASYNC_DISCARD 0x4000ULL /* support async DoM data discard */
848 #define OBD_CONNECT2_ENCRYPT 0x8000ULL /* client-to-disk encrypt */
849 #define OBD_CONNECT2_FIDMAP 0x10000ULL /* FID map */
850 #define OBD_CONNECT2_GETATTR_PFID 0x20000ULL /* pack parent FID in getattr */
851 #define OBD_CONNECT2_LSEEK 0x40000ULL /* SEEK_HOLE/DATA RPC */
852 #define OBD_CONNECT2_DOM_LVB 0x80000ULL /* pack DOM glimpse data in LVB */
853 #define OBD_CONNECT2_REP_MBITS 0x100000ULL /* match reply by mbits, not xid */
854 #define OBD_CONNECT2_ATOMIC_OPEN_LOCK 0x4000000ULL/* request lock on 1st open */
856 * Please DO NOT add flag values here before first ensuring that this same
857 * flag value is not in use on some other branch. Please clear any such
858 * changes with senior engineers before starting to use a new flag. Then,
859 * submit a small patch against EVERY branch that ONLY adds the new flag,
860 * updates obd_connect_names[], adds the flag to check_obd_connect_data(),
861 * and updates wiretests accordingly, so it can be approved and landed easily
862 * to reserve the flag for future use.
865 #define OCD_HAS_FLAG(ocd, flg) \
866 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flg))
869 #ifdef HAVE_LRU_RESIZE_SUPPORT
870 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
872 #define LRU_RESIZE_CONNECT_FLAG 0
875 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
876 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
877 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
878 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
879 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
880 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
881 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
882 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
883 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
884 OBD_CONNECT_EINPROGRESS | \
885 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
886 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
887 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
888 OBD_CONNECT_FLOCK_DEAD | \
889 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
890 OBD_CONNECT_OPEN_BY_FID | \
891 OBD_CONNECT_DIR_STRIPE | OBD_CONNECT_GRANT | \
892 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_SRVLOCK | \
893 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_CKSUM | \
894 OBD_CONNECT_MULTIMODRPCS | \
895 OBD_CONNECT_SUBTREE | OBD_CONNECT_LARGE_ACL | \
896 OBD_CONNECT_GRANT_PARAM | \
897 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
899 #define MDT_CONNECT_SUPPORTED2 (OBD_CONNECT2_FILE_SECCTX | \
900 OBD_CONNECT2_DIR_MIGRATE | \
901 OBD_CONNECT2_SUM_STATFS | \
902 OBD_CONNECT2_OVERSTRIPING | \
904 OBD_CONNECT2_LOCK_CONVERT | \
905 OBD_CONNECT2_ARCHIVE_ID_ARRAY | \
906 OBD_CONNECT2_INC_XID | \
907 OBD_CONNECT2_SELINUX_POLICY | \
908 OBD_CONNECT2_LSOM | \
909 OBD_CONNECT2_ASYNC_DISCARD | \
911 OBD_CONNECT2_CRUSH | \
912 OBD_CONNECT2_ENCRYPT | \
913 OBD_CONNECT2_GETATTR_PFID |\
914 OBD_CONNECT2_LSEEK | OBD_CONNECT2_DOM_LVB |\
915 OBD_CONNECT2_REP_MBITS | \
916 OBD_CONNECT2_ATOMIC_OPEN_LOCK)
918 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
919 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
920 OBD_CONNECT_TRUNCLOCK | OBD_CONNECT_INDEX | \
921 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
922 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
923 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
924 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
925 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
926 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
927 OBD_CONNECT_MAX_EASIZE | \
928 OBD_CONNECT_EINPROGRESS | \
929 OBD_CONNECT_JOBSTATS | \
930 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
931 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
932 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
933 OBD_CONNECT_BULK_MBITS | \
934 OBD_CONNECT_GRANT_PARAM | \
935 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
937 #define OST_CONNECT_SUPPORTED2 (OBD_CONNECT2_LOCKAHEAD | OBD_CONNECT2_INC_XID |\
938 OBD_CONNECT2_ENCRYPT | OBD_CONNECT2_LSEEK |\
939 OBD_CONNECT2_REP_MBITS)
941 #define ECHO_CONNECT_SUPPORTED (OBD_CONNECT_FID | OBD_CONNECT_FLAGS2)
942 #define ECHO_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS
944 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
945 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
946 OBD_CONNECT_PINGLESS |\
947 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_BARRIER | \
950 #define MGS_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS
952 /* Features required for this version of the client to work with server */
953 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_FID | \
954 OBD_CONNECT_ATTRFID | \
957 /* This structure is used for both request and reply.
959 * If we eventually have separate connect data for different types, which we
960 * almost certainly will, then perhaps we stick a union in here. */
961 struct obd_connect_data {
962 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
963 __u32 ocd_version; /* lustre release version number */
964 __u32 ocd_grant; /* initial cache grant amount (bytes) */
965 __u32 ocd_index; /* LOV index to connect to */
966 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
967 __u64 ocd_ibits_known; /* inode bits this client understands */
968 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
969 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
970 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
971 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
972 __u64 ocd_transno; /* first transno from client to be replayed */
973 __u32 ocd_group; /* MDS group on OST */
974 __u32 ocd_cksum_types; /* supported checksum algorithms */
975 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
976 __u32 ocd_instance; /* instance # of this target */
977 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
978 /* Fields after ocd_maxbytes are only accessible by the receiver
979 * if the corresponding flag in ocd_connect_flags is set. Accessing
980 * any field after ocd_maxbytes on the receiver without a valid flag
981 * may result in out-of-bound memory access and kernel oops. */
982 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
983 __u16 padding0; /* added 2.1.0. also fix lustre_swab_connect */
984 __u32 padding1; /* added 2.1.0. also fix lustre_swab_connect */
985 __u64 ocd_connect_flags2;
986 __u64 padding3; /* added 2.1.0. also fix lustre_swab_connect */
987 __u64 padding4; /* added 2.1.0. also fix lustre_swab_connect */
988 __u64 padding5; /* added 2.1.0. also fix lustre_swab_connect */
989 __u64 padding6; /* added 2.1.0. also fix lustre_swab_connect */
990 __u64 padding7; /* added 2.1.0. also fix lustre_swab_connect */
991 __u64 padding8; /* added 2.1.0. also fix lustre_swab_connect */
992 __u64 padding9; /* added 2.1.0. also fix lustre_swab_connect */
993 __u64 paddingA; /* added 2.1.0. also fix lustre_swab_connect */
994 __u64 paddingB; /* added 2.1.0. also fix lustre_swab_connect */
995 __u64 paddingC; /* added 2.1.0. also fix lustre_swab_connect */
996 __u64 paddingD; /* added 2.1.0. also fix lustre_swab_connect */
997 __u64 paddingE; /* added 2.1.0. also fix lustre_swab_connect */
998 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1001 * Please DO NOT use any fields here before first ensuring that this same
1002 * field is not in use on some other branch. Please clear any such changes
1003 * with senior engineers before starting to use a new field. Then, submit
1004 * a small patch against EVERY branch that ONLY adds the new field along with
1005 * the matching OBD_CONNECT flag, so that can be approved and landed easily to
1006 * reserve the flag for future use. */
1009 * Supported checksum algorithms. Up to 32 checksum types are supported.
1010 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1011 * Please update DECLARE_CKSUM_NAME in obd_cksum.h when adding a new
1012 * algorithm and also the OBD_FL_CKSUM* flags, OBD_CKSUM_ALL flag,
1013 * OBD_FL_CKSUM_ALL flag and potentially OBD_CKSUM_T10_ALL flag.
1016 OBD_CKSUM_CRC32 = 0x00000001,
1017 OBD_CKSUM_ADLER = 0x00000002,
1018 OBD_CKSUM_CRC32C = 0x00000004,
1019 OBD_CKSUM_RESERVED = 0x00000008,
1020 OBD_CKSUM_T10IP512 = 0x00000010,
1021 OBD_CKSUM_T10IP4K = 0x00000020,
1022 OBD_CKSUM_T10CRC512 = 0x00000040,
1023 OBD_CKSUM_T10CRC4K = 0x00000080,
1026 #define OBD_CKSUM_T10_ALL (OBD_CKSUM_T10IP512 | OBD_CKSUM_T10IP4K | \
1027 OBD_CKSUM_T10CRC512 | OBD_CKSUM_T10CRC4K)
1029 #define OBD_CKSUM_ALL (OBD_CKSUM_CRC32 | OBD_CKSUM_ADLER | OBD_CKSUM_CRC32C | \
1033 * The default checksum algorithm used on top of T10PI GRD tags for RPC.
1034 * Considering that the checksum-of-checksums is only computing CRC32 on a
1035 * 4KB chunk of GRD tags for a 1MB RPC for 512B sectors, or 16KB of GRD
1036 * tags for 16MB of 4KB sectors, this is only 1/256 or 1/1024 of the
1037 * total data being checksummed, so the checksum type used here should not
1038 * affect overall system performance noticeably.
1040 #define OBD_CKSUM_T10_TOP OBD_CKSUM_ADLER
1043 * OST requests: OBDO & OBD request records
1048 OST_REPLY = 0, /* reply ? */
1064 OST_QUOTACHECK = 18, /* not used since 2.4 */
1066 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1070 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
1072 #define OST_FIRST_OPC OST_REPLY
1075 OBD_FL_INLINEDATA = 0x00000001,
1076 OBD_FL_OBDMDEXISTS = 0x00000002,
1077 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1078 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1079 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1080 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1081 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1082 OBD_FL_NO_PRJQUOTA = 0x00000080, /* the object's project is over
1084 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1085 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1086 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1087 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1088 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1089 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1090 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1091 OBD_FL_CKSUM_T10IP512 = 0x00005000, /* T10PI IP cksum, 512B sector */
1092 OBD_FL_CKSUM_T10IP4K = 0x00006000, /* T10PI IP cksum, 4KB sector */
1093 OBD_FL_CKSUM_T10CRC512 = 0x00007000, /* T10PI CRC cksum, 512B sector */
1094 OBD_FL_CKSUM_T10CRC4K = 0x00008000, /* T10PI CRC cksum, 4KB sector */
1095 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1096 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1097 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1098 * XXX: obsoleted - reserved for old
1099 * clients prior than 2.2 */
1100 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1101 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1102 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1103 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1104 /* OBD_FL_LOCAL_MASK = 0xF0000000, was local-only flags until 2.10 */
1107 * Note that while the original checksum values were separate bits,
1108 * in 2.x we can actually allow all values from 1-31. T10-PI checksum
1109 * types already use values which are not separate bits.
1111 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1112 OBD_FL_CKSUM_CRC32C | OBD_FL_CKSUM_T10IP512 |
1113 OBD_FL_CKSUM_T10IP4K | OBD_FL_CKSUM_T10CRC512 |
1114 OBD_FL_CKSUM_T10CRC4K,
1116 OBD_FL_NO_QUOTA_ALL = OBD_FL_NO_USRQUOTA | OBD_FL_NO_GRPQUOTA |
1121 * All LOV EA magics should have the same postfix, if some new version
1122 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1123 * Lustre, even though the old version system does not recognizes such
1124 * new magic, it still can distinguish the corrupted cases by checking
1125 * the magic's postfix.
1127 #define LOV_MAGIC_MAGIC 0x0BD0
1128 #define LOV_MAGIC_MASK 0xFFFF
1130 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1131 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1132 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1133 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1134 /* reserved for specifying OSTs */
1135 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1136 #define LOV_MAGIC LOV_MAGIC_V1
1137 #define LOV_MAGIC_COMP_V1 (0x0BD60000 | LOV_MAGIC_MAGIC)
1138 #define LOV_MAGIC_FOREIGN (0x0BD70000 | LOV_MAGIC_MAGIC)
1139 #define LOV_MAGIC_SEL (0x0BD80000 | LOV_MAGIC_MAGIC)
1142 * magic for fully defined striping
1143 * the idea is that we should have different magics for striping "hints"
1144 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1145 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1146 * we can't just change it w/o long way preparation, but we still need a
1147 * mechanism to allow LOD to differentiate hint versus ready striping.
1148 * so, at the moment we do a trick: MDT knows what to expect from request
1149 * depending on the case (replay uses ready striping, non-replay req uses
1150 * hints), so MDT replaces magic with appropriate one and now LOD can
1151 * easily understand what's inside -bzzz
1153 * those *_DEF magics are only used on server side internally, they
1154 * won't be put on wire or disk.
1156 #define LOV_MAGIC_DEFINED 0x10000000
1157 #define LOV_MAGIC_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V1)
1158 #define LOV_MAGIC_V3_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V3)
1159 #define LOV_MAGIC_COMP_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_COMP_V1)
1161 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1162 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1164 #define lov_ost_data lov_ost_data_v1
1165 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1166 struct ost_id l_ost_oi; /* OST object ID */
1167 __u32 l_ost_gen; /* generation of this l_ost_idx */
1168 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1171 #define lov_mds_md lov_mds_md_v1
1172 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1173 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1174 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1175 struct ost_id lmm_oi; /* LOV object ID */
1176 __u32 lmm_stripe_size; /* size of stripe in bytes */
1177 /* lmm_stripe_count used to be __u32 */
1178 __u16 lmm_stripe_count; /* num stripes in use for this object */
1179 __u16 lmm_layout_gen; /* layout generation number */
1180 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1183 #define MAX_MD_SIZE_OLD (sizeof(struct lov_mds_md) + \
1184 4 * sizeof(struct lov_ost_data))
1185 #define MAX_MD_SIZE (sizeof(struct lov_comp_md_v1) + \
1186 4 * (sizeof(struct lov_comp_md_entry_v1) + \
1188 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1190 /* This is the default MDT reply size allocated, should the striping be bigger,
1191 * it will be reallocated in mdt_fix_reply.
1192 * 100 stripes is a bit less than 2.5k of data */
1193 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1194 100 * sizeof(struct lov_ost_data))
1196 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1197 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1198 #define XATTR_USER_PREFIX "user."
1199 #define XATTR_TRUSTED_PREFIX "trusted."
1200 #define XATTR_SECURITY_PREFIX "security."
1202 #define XATTR_NAME_SOM "trusted.som"
1203 #define XATTR_NAME_LOV "trusted.lov"
1204 #define XATTR_NAME_LMA "trusted.lma"
1205 #define XATTR_NAME_LMV "trusted.lmv"
1206 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1207 #define XATTR_NAME_LINK "trusted.link"
1208 #define XATTR_NAME_FID "trusted.fid"
1209 #define XATTR_NAME_VERSION "trusted.version"
1210 #define XATTR_NAME_SOM "trusted.som"
1211 #define XATTR_NAME_HSM "trusted.hsm"
1212 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1213 #define XATTR_NAME_DUMMY "trusted.dummy"
1215 #define LL_XATTR_NAME_ENCRYPTION_CONTEXT XATTR_SECURITY_PREFIX"c"
1217 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1218 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1220 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1221 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1222 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1223 struct ost_id lmm_oi; /* LOV object ID */
1224 __u32 lmm_stripe_size; /* size of stripe in bytes */
1225 /* lmm_stripe_count used to be __u32 */
1226 __u16 lmm_stripe_count; /* num stripes in use for this object */
1227 __u16 lmm_layout_gen; /* layout generation number */
1228 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1229 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1232 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1234 if (stripes == (__u16)-1)
1237 if (lmm_magic == LOV_MAGIC_V3)
1238 return sizeof(struct lov_mds_md_v3) +
1239 stripes * sizeof(struct lov_ost_data_v1);
1241 return sizeof(struct lov_mds_md_v1) +
1242 stripes * sizeof(struct lov_ost_data_v1);
1246 lov_mds_md_max_stripe_count(__kernel_size_t buf_size, __u32 lmm_magic)
1248 switch (lmm_magic) {
1249 case LOV_MAGIC_V1: {
1250 struct lov_mds_md_v1 lmm;
1252 if (buf_size < sizeof(lmm))
1255 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1257 case LOV_MAGIC_V3: {
1258 struct lov_mds_md_v3 lmm;
1260 if (buf_size < sizeof(lmm))
1263 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1270 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1271 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1272 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1273 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1274 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1275 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1276 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1277 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1278 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1279 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1280 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1281 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1282 #define OBD_MD_DOM_SIZE (0X00001000ULL) /* Data-on-MDT component size */
1283 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1284 #define OBD_MD_FLPARENT (0x00004000ULL) /* parent FID */
1285 #define OBD_MD_LAYOUT_VERSION (0x00008000ULL) /* OST object layout version */
1286 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1287 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1288 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1289 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1290 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1291 /* OBD_MD_FLQOS (0x00200000ULL) has never been used */
1292 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1293 #define OBD_MD_FLPRJQUOTA (0x00400000ULL) /* over quota flags sent from ost */
1294 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1295 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1296 /* OBD_MD_FLEPOCH (0x04000000ULL) obsolete 2.7.50 */
1297 /* ->mds if epoch opens or closes */
1298 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1299 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1300 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1301 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1302 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1304 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1305 /* OBD_MD_REINT (0x0000000200000000ULL) obsolete 1.8 */
1306 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1307 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1309 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1310 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1311 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1312 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1313 #define OBD_MD_FLAGSTATFS (0x0000010000000000ULL) /* aggregated statfs */
1314 /* OBD_MD_FLMDSCAPA (0x0000020000000000ULL) obsolete 2.7.54 */
1315 /* OBD_MD_FLOSSCAPA (0x0000040000000000ULL) obsolete 2.7.54 */
1316 /* OBD_MD_FLCKSPLIT (0x0000080000000000ULL) obsolete 2.3.58*/
1317 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1318 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1319 * under lock; for xattr
1320 * requests means the
1321 * client holds the lock */
1322 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1324 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1325 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1328 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1329 #define OBD_MD_FLOSTLAYOUT (0x0080000000000000ULL) /* contain ost_layout */
1330 #define OBD_MD_FLPROJID (0x0100000000000000ULL) /* project ID */
1331 #define OBD_MD_SECCTX (0x0200000000000000ULL) /* embed security xattr */
1332 #define OBD_MD_FLLAZYSIZE (0x0400000000000000ULL) /* Lazy size */
1333 #define OBD_MD_FLLAZYBLOCKS (0x0800000000000000ULL) /* Lazy blocks */
1334 #define OBD_MD_FLBTIME (0x1000000000000000ULL) /* birth time */
1335 #define OBD_MD_ENCCTX (0x2000000000000000ULL) /* embed encryption ctx */
1337 #define OBD_MD_FLALLQUOTA (OBD_MD_FLUSRQUOTA | \
1338 OBD_MD_FLGRPQUOTA | \
1341 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1342 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1343 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1344 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1345 OBD_MD_FLPARENT | OBD_MD_FLRDEV | OBD_MD_FLGROUP | \
1346 OBD_MD_FLPROJID | OBD_MD_FLBTIME)
1348 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1350 /* don't forget obdo_fid which is way down at the bottom so it can
1351 * come after the definition of llog_cookie */
1355 HSS_CLEARMASK = 0x02,
1356 HSS_ARCHIVE_ID = 0x04,
1359 struct hsm_state_set {
1361 __u32 hss_archive_id;
1363 __u64 hss_clearmask;
1366 /* ost_body.data values for OST_BRW */
1368 #define OBD_BRW_READ 0x01
1369 #define OBD_BRW_WRITE 0x02
1370 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1371 #define OBD_BRW_NDELAY 0x04 /* Non-delay RPC should be issued for
1372 * this page. Non-delay RPCs have bit
1373 * rq_no_delay set. */
1374 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1375 * transfer and is not accounted in
1377 #define OBD_BRW_CHECK 0x10
1378 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1379 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1380 /* OBD_BRW_NOCACHE is currently neither set nor tested */
1381 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1382 #define OBD_BRW_NOQUOTA 0x100
1383 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1384 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1385 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1386 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1387 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1388 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1389 * that the client is running low on
1390 * space for unstable pages; asking
1391 * it to sync quickly */
1392 #define OBD_BRW_OVER_PRJQUOTA 0x8000 /* Running out of project quota */
1393 #define OBD_BRW_RDMA_ONLY 0x20000 /* RPC contains RDMA-only pages*/
1395 #define OBD_BRW_OVER_ALLQUOTA (OBD_BRW_OVER_USRQUOTA | \
1396 OBD_BRW_OVER_GRPQUOTA | \
1397 OBD_BRW_OVER_PRJQUOTA)
1399 #define OBD_BRW_DONE 0x40000000UL /*
1400 * osd-ldiskfs inernal,
1401 * IO has been issued before
1403 #define OBD_BRW_LOCAL1 0x80000000UL /*
1404 * osd-ldiskfs internal,
1405 * page mapped to real block
1408 #define OBD_BRW_LOCALS (OBD_BRW_LOCAL1 | OBD_BRW_DONE)
1410 #define OBD_MAX_GRANT 0x7fffffffUL /* Max grant allowed to one client: 2 GiB */
1412 #define OBD_OBJECT_EOF LUSTRE_EOF
1414 #define OST_MIN_PRECREATE 32
1415 #define OST_MAX_PRECREATE 20000
1418 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1419 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1420 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1421 * high 16 bits in 2.4 and later */
1422 __u32 ioo_bufcnt; /* number of niobufs for this object */
1425 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1426 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1427 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. */
1428 #define IOOBJ_MAX_BRW_BITS 16
1429 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1430 #define ioobj_max_brw_set(ioo, num) \
1431 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1433 /* multiple of 8 bytes => can array */
1434 struct niobuf_remote {
1440 /* lock value block communicated between the filter and llite */
1442 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1443 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1444 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1445 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1446 #define OST_LVB_IS_ERR(blocks) \
1447 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1448 #define OST_LVB_SET_ERR(blocks, rc) \
1449 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1450 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1473 * lquota data structures
1476 #ifndef QUOTABLOCK_BITS
1477 # define QUOTABLOCK_BITS LUSTRE_QUOTABLOCK_BITS
1480 #ifndef QUOTABLOCK_SIZE
1481 # define QUOTABLOCK_SIZE LUSTRE_QUOTABLOCK_SIZE
1485 # define toqb lustre_stoqb
1488 /* The lquota_id structure is an union of all the possible identifier types that
1489 * can be used with quota, this includes:
1492 * - a FID which can be used for per-directory quota in the future */
1494 struct lu_fid qid_fid; /* FID for per-directory quota */
1495 __u64 qid_uid; /* user identifier */
1496 __u64 qid_gid; /* group identifier */
1497 __u64 qid_projid; /* project identifier */
1500 /* quotactl management */
1501 struct obd_quotactl {
1503 __u32 qc_type; /* see Q_* flag below */
1506 struct obd_dqinfo qc_dqinfo;
1507 struct obd_dqblk qc_dqblk;
1508 char qc_poolname[0];
1511 #define Q_COPY(out, in, member) (out)->member = (in)->member
1513 #define QCTL_COPY(out, in) \
1515 Q_COPY(out, in, qc_cmd); \
1516 Q_COPY(out, in, qc_type); \
1517 Q_COPY(out, in, qc_id); \
1518 Q_COPY(out, in, qc_stat); \
1519 Q_COPY(out, in, qc_dqinfo); \
1520 Q_COPY(out, in, qc_dqblk); \
1521 if (LUSTRE_Q_CMD_IS_POOL(in->qc_cmd)) \
1522 memcpy(out->qc_poolname, \
1524 LOV_MAXPOOLNAME + 1); \
1527 /* Body of quota request used for quota acquire/release RPCs between quota
1528 * master (aka QMT) and slaves (ak QSD). */
1530 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1531 * and type (data or metadata) as well as
1532 * the quota type (user or group). */
1533 union lquota_id qb_id; /* uid or gid or directory FID */
1534 __u32 qb_flags; /* see below */
1536 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1537 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1538 __u64 qb_slv_ver; /* slave index file version */
1539 struct lustre_handle qb_lockh; /* per-ID lock handle */
1540 struct lustre_handle qb_glb_lockh; /* global lock handle */
1541 __u64 qb_padding1[4];
1544 /* When the quota_body is used in the reply of quota global intent
1545 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1546 #define qb_slv_fid qb_fid
1547 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1549 #define qb_qunit qb_usage
1551 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1552 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1553 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1554 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1556 /* Quota types currently supported */
1558 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1559 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1560 LQUOTA_TYPE_PRJ = 0x02, /* maps to PRJQUOTA */
1564 /* There are 2 different resource types on which a quota limit can be enforced:
1565 * - inodes on the MDTs
1566 * - blocks on the OSTs */
1568 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1569 LQUOTA_RES_DT = 0x02,
1571 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1573 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1576 * Space accounting support
1577 * Format of an accounting record, providing disk usage information for a given
1580 struct lquota_acct_rec { /* 16 bytes */
1581 __u64 bspace; /* current space in use */
1582 __u64 ispace; /* current # inodes in use */
1586 * Global quota index support
1587 * Format of a global record, providing global quota settings for a given quota
1590 struct lquota_glb_rec { /* 32 bytes */
1591 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1592 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1593 __u64 qbr_time; /* grace time, in seconds */
1594 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1599 * Slave index support
1600 * Format of a slave record, recording how much space is granted to a given
1603 struct lquota_slv_rec { /* 8 bytes */
1604 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1605 * in #inodes or kbytes */
1608 /* Data structures associated with the quota locks */
1610 /* Glimpse descriptor used for the index & per-ID quota locks */
1611 struct ldlm_gl_lquota_desc {
1612 union lquota_id gl_id; /* quota ID subject to the glimpse */
1613 __u64 gl_flags; /* see LQUOTA_FL* below */
1614 __u64 gl_ver; /* new index version */
1615 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1616 __u64 gl_softlimit; /* new softlimit */
1620 #define gl_qunit gl_hardlimit /* current qunit value used when
1621 * glimpsing per-ID quota locks */
1623 /* quota glimpse flags */
1624 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1626 /* LVB used with quota (global and per-ID) locks */
1628 __u64 lvb_flags; /* see LQUOTA_FL* above */
1629 __u64 lvb_id_may_rel; /* space that might be released later */
1630 __u64 lvb_id_rel; /* space released by the slave for this ID */
1631 __u64 lvb_id_qunit; /* current qunit value */
1635 /* LVB used with global quota lock */
1636 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1644 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1653 MDS_GETATTR_NAME = 34,
1658 MDS_DISCONNECT = 39,
1661 MDS_PIN = 42, /* obsolete, never used in a release */
1662 MDS_UNPIN = 43, /* obsolete, never used in a release */
1664 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1666 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1669 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1671 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1673 MDS_HSM_STATE_GET = 54,
1674 MDS_HSM_STATE_SET = 55,
1675 MDS_HSM_ACTION = 56,
1676 MDS_HSM_PROGRESS = 57,
1677 MDS_HSM_REQUEST = 58,
1678 MDS_HSM_CT_REGISTER = 59,
1679 MDS_HSM_CT_UNREGISTER = 60,
1680 MDS_SWAP_LAYOUTS = 61,
1685 #define MDS_FIRST_OPC MDS_GETATTR
1688 /* opcodes for object update */
1694 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1714 /* the disposition of the intent outlines what was executed */
1715 #define DISP_IT_EXECD 0x00000001
1716 #define DISP_LOOKUP_EXECD 0x00000002
1717 #define DISP_LOOKUP_NEG 0x00000004
1718 #define DISP_LOOKUP_POS 0x00000008
1719 #define DISP_OPEN_CREATE 0x00000010
1720 #define DISP_OPEN_OPEN 0x00000020
1721 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1722 #define DISP_ENQ_OPEN_REF 0x00800000
1723 #define DISP_ENQ_CREATE_REF 0x01000000
1724 #define DISP_OPEN_LOCK 0x02000000
1725 #define DISP_OPEN_LEASE 0x04000000
1726 #define DISP_OPEN_STRIPE 0x08000000
1727 #define DISP_OPEN_DENY 0x10000000
1729 /* INODE LOCK PARTS */
1730 enum mds_ibits_locks {
1731 MDS_INODELOCK_LOOKUP = 0x000001, /* For namespace, dentry etc. Was
1732 * used to protect permission (mode,
1733 * owner, group, etc) before 2.4. */
1734 MDS_INODELOCK_UPDATE = 0x000002, /* size, links, timestamps */
1735 MDS_INODELOCK_OPEN = 0x000004, /* For opened files */
1736 MDS_INODELOCK_LAYOUT = 0x000008, /* for layout */
1738 /* The PERM bit is added in 2.4, and is used to protect permission
1739 * (mode, owner, group, ACL, etc.) separate from LOOKUP lock.
1740 * For remote directories (in DNE) these locks will be granted by
1741 * different MDTs (different LDLM namespace).
1743 * For local directory, the MDT always grants UPDATE|PERM together.
1744 * For remote directory, master MDT (where remote directory is) grants
1745 * UPDATE|PERM, and remote MDT (where name entry is) grants LOOKUP_LOCK.
1747 MDS_INODELOCK_PERM = 0x000010,
1748 MDS_INODELOCK_XATTR = 0x000020, /* non-permission extended attrs */
1749 MDS_INODELOCK_DOM = 0x000040, /* Data for Data-on-MDT files */
1750 /* Do not forget to increase MDS_INODELOCK_NUMBITS when adding bits */
1752 #define MDS_INODELOCK_NUMBITS 7
1753 /* This FULL lock is useful to take on unlink sort of operations */
1754 #define MDS_INODELOCK_FULL ((1 << MDS_INODELOCK_NUMBITS) - 1)
1755 /* DOM lock shouldn't be canceled early, use this macro for ELC */
1756 #define MDS_INODELOCK_ELC (MDS_INODELOCK_FULL & ~MDS_INODELOCK_DOM)
1758 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1759 * but was moved into name[1] along with the OID to avoid consuming the
1760 * name[2,3] fields that need to be used for the quota id (also a FID). */
1762 LUSTRE_RES_ID_SEQ_OFF = 0,
1763 LUSTRE_RES_ID_VER_OID_OFF = 1,
1764 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1765 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1766 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1767 LUSTRE_RES_ID_HSH_OFF = 3
1770 #define MDS_STATUS_CONN 1
1771 #define MDS_STATUS_LOV 2
1774 /* these should be identical to their EXT4_*_FL counterparts, they are
1775 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1776 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1777 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1778 LUSTRE_APPEND_FL = 0x00000020, /* file writes may only append */
1779 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1780 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1781 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1782 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1783 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1784 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1785 LUSTRE_PROJINHERIT_FL = 0x20000000, /* Create with parents projid */
1787 /* These flags will not be identical to any EXT4_*_FL counterparts,
1788 * and only reserved for lustre purpose. Note: these flags might
1789 * be conflict with some of EXT4 flags, so
1790 * 1. these conflict flags needs to be removed when the flag is
1791 * wired by la_flags see osd_attr_get().
1792 * 2. If these flags needs to be stored into inode, they will be
1793 * stored in LMA. see LMAI_XXXX */
1794 LUSTRE_ORPHAN_FL = 0x00002000,
1795 LUSTRE_SET_SYNC_FL = 0x00040000, /* Synchronous setattr on OSTs */
1796 LUSTRE_ENCRYPT_FL = 0x00800000, /* encrypted file */
1798 LUSTRE_LMA_FL_MASKS = LUSTRE_ENCRYPT_FL | LUSTRE_ORPHAN_FL,
1801 #ifndef FS_XFLAG_SYNC
1802 #define FS_XFLAG_SYNC 0x00000020 /* all writes synchronous */
1804 #ifndef FS_XFLAG_NOATIME
1805 #define FS_XFLAG_NOATIME 0x00000040 /* do not update access time */
1807 #ifndef FS_XFLAG_IMMUTABLE
1808 #define FS_XFLAG_IMMUTABLE 0x00000008 /* file cannot be modified */
1810 #ifndef FS_XFLAG_APPEND
1811 #define FS_XFLAG_APPEND 0x00000010 /* all writes append */
1813 #ifndef FS_XFLAG_PROJINHERIT
1814 #define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
1817 /* 64 possible states */
1818 enum md_transient_state {
1819 MS_RESTORE = (1 << 0), /* restore is running */
1823 struct lu_fid mbo_fid1;
1824 struct lu_fid mbo_fid2;
1825 struct lustre_handle mbo_open_handle;
1827 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1831 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1832 __u64 mbo_version; /* was mbo_ioepoch before 2.11 */
1833 __u64 mbo_t_state; /* transient file state defined in
1834 * enum md_transient_state
1835 * was "ino" until 2.4.0 */
1838 __u32 mbo_capability;
1842 __u32 mbo_flags; /* most replies: LUSTRE_*_FL file attributes,
1843 * data_version: OBD_FL_* flags
1846 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1847 __u32 mbo_layout_gen; /* was "generation" until 2.4.0 */
1849 __u32 mbo_eadatasize;
1851 __u32 mbo_max_mdsize;
1852 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1853 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1854 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1856 __u64 mbo_dom_size; /* size of DOM component */
1857 __u64 mbo_dom_blocks; /* blocks consumed by DOM component */
1859 __u64 mbo_padding_9; /* also fix lustre_swab_mdt_body */
1860 __u64 mbo_padding_10;
1863 struct mdt_ioepoch {
1864 struct lustre_handle mio_open_handle;
1865 __u64 mio_unused1; /* was ioepoch */
1866 __u32 mio_unused2; /* was flags */
1870 /* permissions for md_perm.mp_perm */
1872 CFS_SETUID_PERM = 0x01,
1873 CFS_SETGID_PERM = 0x02,
1874 CFS_SETGRP_PERM = 0x04,
1877 struct mdt_rec_setattr {
1887 __u32 sa_padding_1_h;
1888 struct lu_fid sa_fid;
1897 __u32 sa_attr_flags;
1899 __u32 sa_bias; /* some operation flags */
1906 * Attribute flags used in mdt_rec_setattr::sa_valid.
1907 * The kernel's #defines for ATTR_* should not be used over the network
1908 * since the client and MDS may run different kernels (see bug 13828)
1909 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1911 enum mds_attr_flags {
1912 MDS_ATTR_MODE = 0x1ULL, /* = 1 */
1913 MDS_ATTR_UID = 0x2ULL, /* = 2 */
1914 MDS_ATTR_GID = 0x4ULL, /* = 4 */
1915 MDS_ATTR_SIZE = 0x8ULL, /* = 8 */
1916 MDS_ATTR_ATIME = 0x10ULL, /* = 16 */
1917 MDS_ATTR_MTIME = 0x20ULL, /* = 32 */
1918 MDS_ATTR_CTIME = 0x40ULL, /* = 64 */
1919 MDS_ATTR_ATIME_SET = 0x80ULL, /* = 128 */
1920 MDS_ATTR_MTIME_SET = 0x100ULL, /* = 256 */
1921 MDS_ATTR_FORCE = 0x200ULL, /* = 512, change it */
1922 MDS_ATTR_ATTR_FLAG = 0x400ULL, /* = 1024 */
1923 MDS_ATTR_KILL_SUID = 0x800ULL, /* = 2048 */
1924 MDS_ATTR_KILL_SGID = 0x1000ULL, /* = 4096 */
1925 MDS_ATTR_CTIME_SET = 0x2000ULL, /* = 8192 */
1926 MDS_ATTR_FROM_OPEN = 0x4000ULL, /* = 16384, from open O_TRUNC */
1927 MDS_ATTR_BLOCKS = 0x8000ULL, /* = 32768 */
1928 MDS_ATTR_PROJID = 0x10000ULL, /* = 65536 */
1929 MDS_ATTR_LSIZE = 0x20000ULL, /* = 131072 */
1930 MDS_ATTR_LBLOCKS = 0x40000ULL, /* = 262144 */
1931 MDS_ATTR_OVERRIDE = 0x2000000ULL, /* = 33554432 */
1935 /* MDS_CHECK_SPLIT = 1 << 0, obsolete before 2.3.58 */
1936 /* used for remote object getattr/open by name: in the original
1937 * getattr/open request, MDT found the object against name is on another
1938 * MDT, then packed FID and LOOKUP lock in reply and returned -EREMOTE,
1939 * and client knew it's a remote object, then set this flag in
1940 * getattr/open request and sent to the corresponding MDT to finish
1941 * getattr/open, which fetched attributes and UPDATE lock/opened file.
1943 MDS_CROSS_REF = 1 << 1,
1944 /* MDS_VTX_BYPASS = 1 << 2, obsolete since 2.3.54 */
1945 MDS_PERM_BYPASS = 1 << 3,
1946 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1947 MDS_QUOTA_IGNORE = 1 << 5,
1948 /* MDS_CLOSE_CLEANUP = 1 << 6, obsolete since 2.3.51 */
1949 MDS_KEEP_ORPHAN = 1 << 7,
1950 MDS_RECOV_OPEN = 1 << 8,
1951 MDS_DATA_MODIFIED = 1 << 9,
1952 MDS_CREATE_VOLATILE = 1 << 10,
1953 MDS_OWNEROVERRIDE = 1 << 11,
1954 MDS_HSM_RELEASE = 1 << 12,
1955 MDS_CLOSE_MIGRATE = 1 << 13,
1956 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1957 MDS_CLOSE_LAYOUT_MERGE = 1 << 15,
1958 MDS_CLOSE_RESYNC_DONE = 1 << 16,
1959 MDS_CLOSE_LAYOUT_SPLIT = 1 << 17,
1960 MDS_TRUNC_KEEP_LEASE = 1 << 18,
1961 MDS_PCC_ATTACH = 1 << 19,
1962 MDS_CLOSE_UPDATE_TIMES = 1 << 20,
1963 /* setstripe create only, don't restripe if target exists */
1964 MDS_SETSTRIPE_CREATE = 1 << 21,
1967 #define MDS_CLOSE_INTENT (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP | \
1968 MDS_CLOSE_LAYOUT_MERGE | MDS_CLOSE_LAYOUT_SPLIT | \
1969 MDS_CLOSE_RESYNC_DONE)
1971 /* instance of mdt_reint_rec */
1972 struct mdt_rec_create {
1980 __u32 cr_suppgid1_h;
1982 __u32 cr_suppgid2_h;
1983 struct lu_fid cr_fid1;
1984 struct lu_fid cr_fid2;
1985 struct lustre_handle cr_open_handle_old; /* in case of open replay */
1989 __u32 cr_archive_id;
1992 __u64 cr_padding_1; /* rr_blocks */
1995 /* use of helpers set/get_mrc_cr_flags() is needed to access
1996 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
1997 * extend cr_flags size without breaking 1.8 compat */
1998 __u32 cr_flags_l; /* for use with open, low 32 bits */
1999 __u32 cr_flags_h; /* for use with open, high 32 bits */
2000 __u32 cr_umask; /* umask for create */
2001 __u32 cr_padding_4; /* rr_padding_4 */
2004 /* instance of mdt_reint_rec */
2005 struct mdt_rec_link {
2013 __u32 lk_suppgid1_h;
2015 __u32 lk_suppgid2_h;
2016 struct lu_fid lk_fid1;
2017 struct lu_fid lk_fid2;
2019 __u64 lk_padding_1; /* rr_atime */
2020 __u64 lk_padding_2; /* rr_ctime */
2021 __u64 lk_padding_3; /* rr_size */
2022 __u64 lk_padding_4; /* rr_blocks */
2024 __u32 lk_padding_5; /* rr_mode */
2025 __u32 lk_padding_6; /* rr_flags */
2026 __u32 lk_padding_7; /* rr_padding_2 */
2027 __u32 lk_padding_8; /* rr_padding_3 */
2028 __u32 lk_padding_9; /* rr_padding_4 */
2031 /* instance of mdt_reint_rec */
2032 struct mdt_rec_unlink {
2040 __u32 ul_suppgid1_h;
2042 __u32 ul_suppgid2_h;
2043 struct lu_fid ul_fid1;
2044 struct lu_fid ul_fid2;
2046 __u64 ul_padding_2; /* rr_atime */
2047 __u64 ul_padding_3; /* rr_ctime */
2048 __u64 ul_padding_4; /* rr_size */
2049 __u64 ul_padding_5; /* rr_blocks */
2052 __u32 ul_padding_6; /* rr_flags */
2053 __u32 ul_padding_7; /* rr_padding_2 */
2054 __u32 ul_padding_8; /* rr_padding_3 */
2055 __u32 ul_padding_9; /* rr_padding_4 */
2058 /* instance of mdt_reint_rec */
2059 struct mdt_rec_rename {
2067 __u32 rn_suppgid1_h;
2069 __u32 rn_suppgid2_h;
2070 struct lu_fid rn_fid1;
2071 struct lu_fid rn_fid2;
2073 __u64 rn_padding_1; /* rr_atime */
2074 __u64 rn_padding_2; /* rr_ctime */
2075 __u64 rn_padding_3; /* rr_size */
2076 __u64 rn_padding_4; /* rr_blocks */
2077 __u32 rn_bias; /* some operation flags */
2078 __u32 rn_mode; /* cross-ref rename has mode */
2079 __u32 rn_padding_5; /* rr_flags */
2080 __u32 rn_padding_6; /* rr_padding_2 */
2081 __u32 rn_padding_7; /* rr_padding_3 */
2082 __u32 rn_padding_8; /* rr_padding_4 */
2085 /* instance of mdt_reint_rec */
2086 struct mdt_rec_setxattr {
2094 __u32 sx_suppgid1_h;
2096 __u32 sx_suppgid2_h;
2097 struct lu_fid sx_fid;
2098 __u64 sx_padding_1; /* These three are rr_fid2 */
2103 __u64 sx_padding_5; /* rr_ctime */
2104 __u64 sx_padding_6; /* rr_size */
2105 __u64 sx_padding_7; /* rr_blocks */
2108 __u32 sx_padding_8; /* rr_flags */
2109 __u32 sx_padding_9; /* rr_padding_2 */
2110 __u32 sx_padding_10; /* rr_padding_3 */
2111 __u32 sx_padding_11; /* rr_padding_4 */
2114 /* instance of mdt_reint_rec
2115 * FLR: for file resync MDS_REINT_RESYNC RPC. */
2116 struct mdt_rec_resync {
2124 __u32 rs_suppgid1_h;
2126 __u32 rs_suppgid2_h;
2127 struct lu_fid rs_fid;
2128 __u8 rs_padding0[sizeof(struct lu_fid)];
2129 struct lustre_handle rs_lease_handle; /* rr_mtime */
2130 __s64 rs_padding1; /* rr_atime */
2131 __s64 rs_padding2; /* rr_ctime */
2132 __u64 rs_padding3; /* rr_size */
2133 __u64 rs_padding4; /* rr_blocks */
2135 __u32 rs_padding5; /* rr_mode */
2136 __u32 rs_padding6; /* rr_flags */
2137 __u32 rs_padding7; /* rr_flags_h */
2138 __u32 rs_padding8; /* rr_umask */
2140 __u16 rs_padding9; /* rr_padding_4 */
2144 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2145 * Do NOT change the size of various members, otherwise the value
2146 * will be broken in lustre_swab_mdt_rec_reint().
2148 * If you add new members in other mdt_reint_xxx structres and need to use the
2149 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2151 struct mdt_rec_reint {
2159 __u32 rr_suppgid1_h;
2161 __u32 rr_suppgid2_h;
2162 struct lu_fid rr_fid1;
2163 struct lu_fid rr_fid2;
2175 __u16 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2178 #define LMV_DESC_QOS_MAXAGE_DEFAULT 60 /* Seconds */
2180 /* lmv structures */
2182 __u32 ld_tgt_count; /* how many MDS's */
2183 __u32 ld_active_tgt_count; /* how many active */
2184 __u32 ld_default_stripe_count; /* how many objects are used */
2185 __u32 ld_pattern; /* default hash pattern */
2186 __u64 ld_default_hash_size;
2187 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2188 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2189 __u32 ld_qos_maxage; /* in second */
2190 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2191 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2192 struct obd_uuid ld_uuid;
2195 /* LMV layout EA, and it will be stored both in master and slave object */
2196 struct lmv_mds_md_v1 {
2198 __u32 lmv_stripe_count;
2199 __u32 lmv_master_mdt_index; /* On master object, it is master
2200 * MDT index, on slave object, it
2201 * is stripe index of the slave obj */
2202 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2203 * which hash function to be used,
2204 * Note: only lower 16 bits is being
2205 * used for now. Higher 16 bits will
2206 * be used to mark the object status,
2207 * for example migrating or dead. */
2208 __u32 lmv_layout_version; /* increased each time layout changed,
2209 * by directory migration, restripe
2211 __u32 lmv_migrate_offset; /* once this is set, it means this
2212 * directory is been migrated, stripes
2213 * before this offset belong to target,
2214 * from this to source. */
2215 __u32 lmv_migrate_hash; /* hash type of source stripes of
2216 * migrating directory */
2219 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
2220 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2223 /* stripe count before directory split */
2224 #define lmv_split_offset lmv_migrate_offset
2225 /* stripe count after directory merge */
2226 #define lmv_merge_offset lmv_migrate_offset
2227 /* directory hash type after merge */
2228 #define lmv_merge_hash lmv_migrate_hash
2230 /* foreign LMV EA */
2231 struct lmv_foreign_md {
2232 __u32 lfm_magic; /* magic number = LMV_MAGIC_FOREIGN */
2233 __u32 lfm_length; /* length of lfm_value */
2234 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
2235 __u32 lfm_flags; /* flags, type specific */
2236 char lfm_value[]; /* free format value */
2239 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2240 #define LMV_MAGIC LMV_MAGIC_V1
2242 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2243 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2244 #define LMV_MAGIC_FOREIGN 0x0CD50CD0 /* magic for lmv foreign */
2247 * The FNV-1a hash algorithm is as follows:
2248 * hash = FNV_offset_basis
2249 * for each octet_of_data to be hashed
2250 * hash = hash XOR octet_of_data
2251 * hash = hash × FNV_prime
2253 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2255 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2256 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2258 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2259 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2260 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, __kernel_size_t size)
2262 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2263 const unsigned char *p = buf;
2266 for (i = 0; i < size; i++) {
2268 hash *= LUSTRE_FNV_1A_64_PRIME;
2274 /* CRUSH placement group count */
2275 #define LMV_CRUSH_PG_COUNT 4096
2279 struct lmv_mds_md_v1 lmv_md_v1;
2280 struct lmv_user_md lmv_user_md;
2281 struct lmv_foreign_md lmv_foreign_md;
2284 static inline __kernel_ssize_t lmv_mds_md_size(int stripe_count,
2285 unsigned int lmm_magic)
2287 __kernel_ssize_t len = -EINVAL;
2289 switch (lmm_magic) {
2290 case LMV_MAGIC_V1: {
2291 struct lmv_mds_md_v1 *lmm1;
2293 len = sizeof(*lmm1);
2294 len += stripe_count * sizeof(lmm1->lmv_stripe_fids[0]);
2302 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2304 switch (__le32_to_cpu(lmm->lmv_magic)) {
2306 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2307 case LMV_USER_MAGIC:
2308 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2314 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2316 switch (__le32_to_cpu(lmm->lmv_magic)) {
2318 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2319 case LMV_USER_MAGIC:
2320 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2330 FLD_FIRST_OPC = FLD_QUERY
2336 SEQ_FIRST_OPC = SEQ_QUERY
2340 SEQ_ALLOC_SUPER = 0,
2352 LFSCK_NOTIFY = 1101,
2355 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2359 * LOV data structures
2362 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2363 /* The size of the buffer the lov/mdc reserves for the
2364 * array of UUIDs returned by the MDS. With the current
2365 * protocol, this will limit the max number of OSTs per LOV */
2367 #define LOV_DESC_MAGIC 0xB0CCDE5C
2368 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2369 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2371 /* LOV settings descriptor (should only contain static info) */
2373 __u32 ld_tgt_count; /* how many OBD's */
2374 __u32 ld_active_tgt_count; /* how many active */
2375 __s32 ld_default_stripe_count; /* how many objects are used */
2376 __u32 ld_pattern; /* default PATTERN_RAID0 */
2377 __u64 ld_default_stripe_size; /* in bytes */
2378 __s64 ld_default_stripe_offset; /* starting OST index */
2379 __u32 ld_padding_0; /* unused */
2380 __u32 ld_qos_maxage; /* in second */
2381 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2382 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2383 struct obd_uuid ld_uuid;
2386 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2391 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2396 LDLM_BL_CALLBACK = 104,
2397 LDLM_CP_CALLBACK = 105,
2398 LDLM_GL_CALLBACK = 106,
2399 LDLM_SET_INFO = 107,
2402 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2404 #define RES_NAME_SIZE 4
2405 struct ldlm_res_id {
2406 __u64 name[RES_NAME_SIZE];
2409 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2410 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2411 (unsigned long long)(res)->lr_name.name[1], \
2412 (unsigned long long)(res)->lr_name.name[2], \
2413 (unsigned long long)(res)->lr_name.name[3]
2429 #define LCK_MODE_NUM 8
2439 #define LDLM_MIN_TYPE LDLM_PLAIN
2441 struct ldlm_extent {
2447 static inline bool ldlm_extent_equal(const struct ldlm_extent *ex1,
2448 const struct ldlm_extent *ex2)
2450 return ex1->start == ex2->start && ex1->end == ex2->end;
2453 struct ldlm_inodebits {
2456 __u64 try_bits; /* optional bits to try */
2457 __u64 cancel_bits; /* for lock convert */
2462 struct ldlm_flock_wire {
2470 /* it's important that the fields of the ldlm_extent structure match
2471 * the first fields of the ldlm_flock structure because there is only
2472 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2473 * this ever changes we will need to swab the union differently based
2474 * on the resource type. */
2476 union ldlm_wire_policy_data {
2477 struct ldlm_extent l_extent;
2478 struct ldlm_flock_wire l_flock;
2479 struct ldlm_inodebits l_inodebits;
2482 struct barrier_lvb {
2488 struct ldlm_gl_barrier_desc {
2494 union ldlm_gl_desc {
2495 struct ldlm_gl_lquota_desc lquota_desc;
2496 struct ldlm_gl_barrier_desc barrier_desc;
2499 enum ldlm_intent_flags {
2500 IT_OPEN = 0x00000001,
2501 IT_CREAT = 0x00000002,
2502 IT_OPEN_CREAT = IT_OPEN | IT_CREAT, /* To allow case label. */
2503 IT_READDIR = 0x00000004, /* Used by mdc, not put on the wire. */
2504 IT_GETATTR = 0x00000008,
2505 IT_LOOKUP = 0x00000010,
2506 /* IT_UNLINK = 0x00000020, Obsolete. */
2507 /* IT_TRUNC = 0x00000040, Obsolete. */
2508 IT_GETXATTR = 0x00000080,
2509 /* IT_EXEC = 0x00000100, Obsolete. */
2510 /* IT_PIN = 0x00000200, Obsolete. */
2511 IT_LAYOUT = 0x00000400,
2512 IT_QUOTA_DQACQ = 0x00000800,
2513 IT_QUOTA_CONN = 0x00001000,
2514 /* IT_SETXATTR = 0x00002000, Obsolete. */
2515 IT_GLIMPSE = 0x00004000,
2516 IT_BRW = 0x00008000,
2519 struct ldlm_intent {
2523 struct ldlm_resource_desc {
2524 enum ldlm_type lr_type;
2525 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2526 struct ldlm_res_id lr_name;
2529 struct ldlm_lock_desc {
2530 struct ldlm_resource_desc l_resource;
2531 enum ldlm_mode l_req_mode;
2532 enum ldlm_mode l_granted_mode;
2533 union ldlm_wire_policy_data l_policy_data;
2536 #define LDLM_LOCKREQ_HANDLES 2
2537 #define LDLM_ENQUEUE_CANCEL_OFF 1
2539 struct ldlm_request {
2540 __u32 lock_flags; /* LDLM_FL_*, see lustre_dlm_flags.h */
2541 __u32 lock_count; /* number of locks in lock_handle[] */
2542 struct ldlm_lock_desc lock_desc;/* lock descriptor */
2543 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2548 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2549 struct ldlm_lock_desc lock_desc;
2550 struct lustre_handle lock_handle;
2551 __u64 lock_policy_res1;
2552 __u64 lock_policy_res2;
2555 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2556 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2559 * Opcodes for mountconf (mgs and mgc)
2563 MGS_DISCONNECT = 251,
2564 MGS_EXCEPTION = 252, /* node died, etc. */
2565 MGS_TARGET_REG = 253, /* whenever target starts up */
2566 MGS_TARGET_DEL = 254,
2568 MGS_CONFIG_READ = 256,
2570 MGS_FIRST_OPC = MGS_CONNECT
2573 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 18, 53, 0)
2574 #define MGS_PARAM_MAXLEN 1024
2575 #define KEY_SET_INFO "set_info"
2577 struct mgs_send_param {
2578 char mgs_param[MGS_PARAM_MAXLEN];
2582 /* We pass this info to the MGS so it can write config logs */
2583 #define MTI_NAME_MAXLEN 64
2584 #define MTI_PARAM_MAXLEN 4096
2585 #define MTI_NIDS_MAX 32
2586 struct mgs_target_info {
2587 __u32 mti_lustre_ver;
2588 __u32 mti_stripe_index;
2589 __u32 mti_config_ver;
2590 __u32 mti_flags; /* LDD_F_* */
2591 __u32 mti_nid_count;
2592 __u32 mti_instance; /* Running instance of target */
2593 char mti_fsname[MTI_NAME_MAXLEN];
2594 char mti_svname[MTI_NAME_MAXLEN];
2595 char mti_uuid[sizeof(struct obd_uuid)];
2596 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t) */
2597 char mti_params[MTI_PARAM_MAXLEN];
2600 struct mgs_nidtbl_entry {
2601 __u64 mne_version; /* table version of this entry */
2602 __u32 mne_instance; /* target instance # */
2603 __u32 mne_index; /* target index */
2604 __u32 mne_length; /* length of this entry - by bytes */
2605 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2606 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2607 __u8 mne_nid_size; /* size of each NID, by bytes */
2608 __u8 mne_nid_count; /* # of NIDs in buffer */
2610 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2615 MGS_CFG_T_CONFIG = 0,
2616 MGS_CFG_T_SPTLRPC = 1,
2617 MGS_CFG_T_RECOVER = 2,
2618 MGS_CFG_T_PARAMS = 3,
2619 MGS_CFG_T_NODEMAP = 4,
2620 MGS_CFG_T_BARRIER = 5,
2624 struct mgs_config_body {
2625 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2626 __u64 mcb_offset; /* next index of config log to request */
2627 __u16 mcb_type; /* type of log: MGS_CFG_T_[CONFIG|RECOVER] */
2628 __u8 mcb_nm_cur_pass;
2629 __u8 mcb_bits; /* bits unit size of config log */
2630 __u32 mcb_units; /* # of units for bulk transfer */
2633 struct mgs_config_res {
2634 __u64 mcr_offset; /* index of last config log */
2636 __u64 mcr_size; /* size of the log */
2637 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2641 /* Config marker flags (in config log) */
2642 #define CM_START 0x01
2644 #define CM_SKIP 0x04
2645 #define CM_UPGRADE146 0x08
2646 #define CM_EXCLUDE 0x10
2647 #define CM_START_SKIP (CM_START | CM_SKIP)
2650 __u32 cm_step; /* aka config version */
2652 __u32 cm_vers; /* lustre release version number */
2653 __u32 cm_padding; /* 64 bit align */
2654 __s64 cm_createtime; /*when this record was first created */
2655 __s64 cm_canceltime; /*when this record is no longer valid*/
2656 char cm_tgtname[MTI_NAME_MAXLEN];
2657 char cm_comment[MTI_NAME_MAXLEN];
2661 * Opcodes for multiple servers.
2665 /* OBD_LOG_CANCEL = 401, obsolete since 1.5 */
2666 /* OBD_QC_CALLBACK = 402, obsolete since 2.4 */
2669 OBD_FIRST_OPC = OBD_PING
2673 * llog contexts indices.
2675 * There is compatibility problem with indexes below, they are not
2676 * continuous and must keep their numbers for compatibility needs.
2677 * See LU-5218 for details.
2680 LLOG_CONFIG_ORIG_CTXT = 0,
2681 LLOG_CONFIG_REPL_CTXT = 1,
2682 LLOG_MDS_OST_ORIG_CTXT = 2,
2683 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2684 LLOG_SIZE_ORIG_CTXT = 4,
2685 LLOG_SIZE_REPL_CTXT = 5,
2686 LLOG_TEST_ORIG_CTXT = 8,
2687 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2688 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2689 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2690 /* for multiple changelog consumers */
2691 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2692 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2693 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log */
2694 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log */
2698 /** Identifier for a single log object */
2700 struct ost_id lgl_oi;
2702 } __attribute__((packed));
2704 /** Records written to the CATALOGS list */
2705 #define CATLIST "CATALOGS"
2707 struct llog_logid lci_logid;
2711 } __attribute__((packed));
2713 /* Log data record types - there is no specific reason that these need to
2714 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2716 #define LLOG_OP_MAGIC 0x10600000
2717 #define LLOG_OP_MASK 0xfff00000
2720 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2721 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2722 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2723 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2724 REINT_UNLINK, /* obsolete after 2.5.0 */
2725 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2727 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2728 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2730 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2731 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2732 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2733 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2734 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2735 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2736 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2737 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000,
2738 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2739 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2742 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2743 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2745 /** Log record header - stored in little endian order.
2746 * Each record must start with this struct, end with a llog_rec_tail,
2747 * and be a multiple of 256 bits in size.
2749 struct llog_rec_hdr {
2754 } __attribute__((packed));
2756 struct llog_rec_tail {
2759 } __attribute__((packed));
2761 /* Where data follow just after header */
2762 #define REC_DATA(ptr) \
2763 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2765 #define REC_DATA_LEN(rec) \
2766 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2767 sizeof(struct llog_rec_tail))
2769 struct llog_logid_rec {
2770 struct llog_rec_hdr lid_hdr;
2771 struct llog_logid lid_id;
2775 struct llog_rec_tail lid_tail;
2776 } __attribute__((packed));
2778 struct llog_unlink_rec {
2779 struct llog_rec_hdr lur_hdr;
2783 struct llog_rec_tail lur_tail;
2784 } __attribute__((packed));
2786 struct llog_unlink64_rec {
2787 struct llog_rec_hdr lur_hdr;
2788 struct lu_fid lur_fid;
2789 __u32 lur_count; /* to destroy the lost precreated */
2793 struct llog_rec_tail lur_tail;
2794 } __attribute__((packed));
2796 struct llog_setattr64_rec {
2797 struct llog_rec_hdr lsr_hdr;
2798 struct ost_id lsr_oi;
2804 struct llog_rec_tail lsr_tail;
2805 } __attribute__((packed));
2807 /* Extended to support project quota */
2808 struct llog_setattr64_rec_v2 {
2809 struct llog_rec_hdr lsr_hdr;
2810 struct ost_id lsr_oi;
2817 __u32 lsr_layout_version;
2820 struct llog_rec_tail lsr_tail;
2821 } __attribute__((packed));
2823 struct llog_size_change_rec {
2824 struct llog_rec_hdr lsc_hdr;
2825 struct ll_fid lsc_fid;
2830 struct llog_rec_tail lsc_tail;
2831 } __attribute__((packed));
2833 #define CHANGELOG_MAGIC 0xca103000
2835 /** \a changelog_rec_type's that can't be masked */
2836 #define CHANGELOG_MINMASK (1 << CL_MARK)
2837 /** bits covering all \a changelog_rec_type's */
2838 #define CHANGELOG_ALLMASK 0XFFFFFFFF
2839 /** default \a changelog_rec_type mask. Allow all of them, except
2840 * CL_ATIME since it can really be time consuming, and not necessary
2842 * Remove also CL_OPEN, CL_GETXATTR and CL_DN_OPEN from default list as it can
2843 * be costly and only necessary for audit purpose.
2845 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & \
2846 ~(1 << CL_ATIME | 1 << CL_OPEN | 1 << CL_GETXATTR | \
2849 /* changelog llog name, needed by client replicators */
2850 #define CHANGELOG_CATALOG "changelog_catalog"
2852 struct changelog_setinfo {
2855 } __attribute__((packed));
2857 /** changelog record */
2858 struct llog_changelog_rec {
2859 struct llog_rec_hdr cr_hdr;
2860 struct changelog_rec cr; /**< Variable length field */
2861 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2862 } __attribute__((packed));
2864 #define CHANGELOG_USER_PREFIX "cl"
2866 struct llog_changelog_user_rec {
2867 struct llog_rec_hdr cur_hdr;
2869 /* only intended to be used in relative time comparisons to
2870 * detect idle users */
2873 struct llog_rec_tail cur_tail;
2874 } __attribute__((packed));
2876 enum agent_req_status {
2884 static inline const char *agent_req_status2name(enum agent_req_status ars)
2902 struct llog_agent_req_rec {
2903 struct llog_rec_hdr arr_hdr; /**< record header */
2904 __u32 arr_status; /**< status of the request */
2906 * agent_req_status */
2907 __u32 arr_archive_id; /**< backend archive number */
2908 __u64 arr_flags; /**< req flags */
2909 __u64 arr_compound_id; /**< compound cookie, ignored */
2910 __u64 arr_req_create; /**< req. creation time */
2911 __u64 arr_req_change; /**< req. status change time */
2912 struct hsm_action_item arr_hai; /**< req. to the agent */
2913 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2914 } __attribute__((packed));
2916 /* Old llog gen for compatibility */
2920 } __attribute__((packed));
2922 struct llog_gen_rec {
2923 struct llog_rec_hdr lgr_hdr;
2924 struct llog_gen lgr_gen;
2928 struct llog_rec_tail lgr_tail;
2931 /* flags for the logs */
2933 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2934 LLOG_F_IS_CAT = 0x2,
2935 LLOG_F_IS_PLAIN = 0x4,
2936 LLOG_F_EXT_JOBID = 0x8,
2937 LLOG_F_IS_FIXSIZE = 0x10,
2938 LLOG_F_EXT_EXTRA_FLAGS = 0x20,
2939 LLOG_F_EXT_X_UIDGID = 0x40,
2940 LLOG_F_EXT_X_NID = 0x80,
2941 LLOG_F_EXT_X_OMODE = 0x100,
2942 LLOG_F_EXT_X_XATTR = 0x200,
2943 LLOG_F_RM_ON_ERR = 0x400,
2945 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
2946 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
2947 * because the catlog record is usually fixed size, but its plain
2948 * log record can be variable */
2949 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID | LLOG_F_EXT_EXTRA_FLAGS |
2950 LLOG_F_EXT_X_UIDGID | LLOG_F_EXT_X_NID |
2951 LLOG_F_EXT_X_OMODE | LLOG_F_EXT_X_XATTR,
2954 /* means first record of catalog */
2956 LLOG_CAT_FIRST = -1,
2959 /* On-disk header structure of each log object, stored in little endian order */
2960 #define LLOG_MIN_CHUNK_SIZE 8192
2961 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
2962 * - sizeof(llh_bitmap) */
2963 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
2964 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
2966 struct llog_log_hdr {
2967 struct llog_rec_hdr llh_hdr;
2968 __s64 llh_timestamp;
2970 __u32 llh_bitmap_offset;
2973 /* for a catalog the first/oldest and still in-use plain slot is just
2974 * next to it. It will serve as the upper limit after Catalog has
2977 struct obd_uuid llh_tgtuuid;
2978 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
2979 /* These fields must always be at the end of the llog_log_hdr.
2980 * Note: llh_bitmap size is variable because llog chunk size could be
2981 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
2982 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
2983 * llh_tail should only be refered by LLOG_HDR_TAIL().
2984 * But this structure is also used by client/server llog interface
2985 * (see llog_client.c), it will be kept in its original way to avoid
2986 * compatiblity issue. */
2987 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
2988 struct llog_rec_tail llh_tail;
2989 } __attribute__((packed));
2990 #undef LLOG_HEADER_SIZE
2991 #undef LLOG_BITMAP_BYTES
2993 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
2994 llh->llh_bitmap_offset - \
2995 sizeof(llh->llh_tail)) * 8)
2996 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
2997 (llh)->llh_bitmap_offset)
2998 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
2999 llh->llh_hdr.lrh_len - \
3000 sizeof(llh->llh_tail)))
3002 /** log cookies are used to reference a specific log file and a record therein,
3003 and pass record offset from llog_process_thread to llog_write */
3004 struct llog_cookie {
3006 struct llog_logid lgc_lgl;
3012 } __attribute__((packed));
3014 /** llog protocol */
3015 enum llogd_rpc_ops {
3016 LLOG_ORIGIN_HANDLE_CREATE = 501,
3017 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
3018 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
3019 /* LLOG_ORIGIN_HANDLE_WRITE_REC = 504, Obsolete by 2.1. */
3020 /* LLOG_ORIGIN_HANDLE_CLOSE = 505, Obsolete by 1.8. */
3021 /* LLOG_ORIGIN_CONNECT = 506, Obsolete by 2.4. */
3022 /* LLOG_CATINFO = 507, Obsolete by 2.3. */
3023 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
3024 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* Obsolete by 2.11. */
3026 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
3030 struct llog_logid lgd_logid;
3032 __u32 lgd_llh_flags;
3034 __u32 lgd_saved_index;
3036 __u64 lgd_cur_offset;
3037 } __attribute__((packed));
3039 struct llogd_conn_body {
3040 struct llog_gen lgdc_gen;
3041 struct llog_logid lgdc_logid;
3042 __u32 lgdc_ctxt_idx;
3043 } __attribute__((packed));
3045 /* Note: 64-bit types are 64-bit aligned in structure */
3047 __u64 o_valid; /* hot fields in this obdo */
3050 __u64 o_size; /* o_size-o_blocks == ost_lvb */
3054 __u64 o_blocks; /* brw: cli sent cached bytes */
3057 /* 32-bit fields start here: keep an even number of them via padding */
3058 __u32 o_blksize; /* optimal IO blocksize */
3059 __u32 o_mode; /* brw: cli sent cache remain */
3063 __u32 o_nlink; /* brw: checksum */
3065 __u32 o_misc; /* brw: o_dropped */
3067 __u64 o_ioepoch; /* epoch in ost writes */
3068 __u32 o_stripe_idx; /* holds stripe idx */
3070 struct lustre_handle o_handle; /* brw: lock handle to prolong
3072 /* Originally, the field is llog_cookie for destroy with unlink cookie
3073 * from MDS, it is obsolete in 2.8. Then reuse it by client to transfer
3074 * layout and PFL information in IO, setattr RPCs. Since llog_cookie is
3075 * not used on wire any longer, remove it from the obdo, then it can be
3076 * enlarged freely in the further without affect related RPCs.
3078 * sizeof(ost_layout) + sieof(__u32) == sizeof(llog_cookie). */
3079 struct ost_layout o_layout;
3080 __u32 o_layout_version;
3084 __u64 o_data_version; /* getattr: sum of iversion for
3086 * brw: grant space consumed on
3087 * the client for the write */
3089 __u32 o_padding_4; /* also fix
3090 * lustre_swab_obdo() */
3095 #define o_dirty o_blocks
3096 #define o_undirty o_mode
3097 #define o_dropped o_misc
3098 #define o_cksum o_nlink
3099 #define o_grant_used o_data_version
3100 #define o_falloc_mode o_nlink
3102 struct lfsck_request {
3114 __u16 lr_async_windows;
3116 struct lu_fid lr_fid;
3117 struct lu_fid lr_fid2;
3125 struct lfsck_reply {
3132 LE_LASTID_REBUILDING = 1,
3133 LE_LASTID_REBUILT = 2,
3139 /* LE_FID_ACCESSED = 8, moved to lfsck_events_local */
3141 LE_CONDITIONAL_DESTROY = 10,
3142 LE_PAIRS_VERIFY = 11,
3143 LE_SET_LMV_MASTER = 15,
3144 LE_SET_LMV_SLAVE = 16,
3147 enum lfsck_event_flags {
3148 LEF_TO_OST = 0x00000001,
3149 LEF_FROM_OST = 0x00000002,
3150 LEF_SET_LMV_HASH = 0x00000004,
3151 LEF_SET_LMV_ALL = 0x00000008,
3152 LEF_RECHECK_NAME_HASH = 0x00000010,
3153 LEF_QUERY_ALL = 0x00000020,
3156 /* request structure for OST's */
3161 /* Key for FIEMAP to be used in get_info calls */
3162 struct ll_fiemap_info_key {
3164 struct obdo lfik_oa;
3165 struct fiemap lfik_fiemap;
3168 #define IDX_INFO_MAGIC 0x3D37CC37
3170 /* Index file transfer through the network. The server serializes the index into
3171 * a byte stream which is sent to the client via a bulk transfer */
3175 /* reply: see idx_info_flags below */
3178 /* request & reply: number of lu_idxpage (to be) transferred */
3182 /* request: requested attributes passed down to the iterator API */
3185 /* request & reply: index file identifier (FID) */
3186 struct lu_fid ii_fid;
3188 /* reply: version of the index file before starting to walk the index.
3189 * Please note that the version can be modified at any time during the
3193 /* request: hash to start with:
3194 * reply: hash of the first entry of the first lu_idxpage and hash
3195 * of the entry to read next if any */
3196 __u64 ii_hash_start;
3199 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3203 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3212 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3214 /* List of flags used in idx_info::ii_flags */
3215 enum idx_info_flags {
3216 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3217 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3218 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3219 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3220 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
3223 #define LIP_MAGIC 0x8A6D6B6C
3225 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3227 /* 16-byte header */
3230 __u16 lip_nr; /* number of entries in the container */
3231 __u64 lip_pad0; /* additional padding for future use */
3233 /* key/record pairs are stored in the remaining 4080 bytes.
3234 * depending upon the flags in idx_info::ii_flags, each key/record
3235 * pair might be preceded by:
3237 * - the key size (II_FL_VARKEY is set)
3238 * - the record size (II_FL_VARREC is set)
3240 * For the time being, we only support fixed-size key & record. */
3241 char lip_entries[0];
3244 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3246 /* Gather all possible type associated with a 4KB container */
3248 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3249 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3250 char lp_array[LU_PAGE_SIZE];
3253 /* security opcodes */
3256 SEC_CTX_INIT_CONT = 802,
3259 SEC_FIRST_OPC = SEC_CTX_INIT
3262 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3263 #define LINK_EA_MAGIC 0x11EAF1DFUL
3264 struct link_ea_header {
3267 __u64 leh_len; /* total size */
3268 __u32 leh_overflow_time;
3272 /** Hardlink data is name and parent fid.
3273 * Stored in this crazy struct for maximum packing and endian-neutrality
3275 struct link_ea_entry {
3276 /** __u16 stored big-endian, unaligned */
3277 unsigned char lee_reclen[2];
3278 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3280 } __attribute__((packed));
3282 /** fid2path request/reply structure */
3283 struct getinfo_fid2path {
3284 struct lu_fid gf_fid;
3290 struct lu_fid gf_root_fid[0];
3292 } __attribute__((packed));
3294 /** path2parent request/reply structures */
3296 struct lu_fid gp_fid; /**< parent FID */
3297 __u32 gp_linkno; /**< hardlink number */
3298 __u32 gp_name_size; /**< size of the name field */
3299 char gp_name[0]; /**< zero-terminated link name */
3300 } __attribute__((packed));
3302 enum layout_intent_opc {
3303 LAYOUT_INTENT_ACCESS = 0, /** generic access */
3304 LAYOUT_INTENT_READ = 1, /** not used */
3305 LAYOUT_INTENT_WRITE = 2, /** write file, for comp layout */
3306 LAYOUT_INTENT_GLIMPSE = 3, /** not used */
3307 LAYOUT_INTENT_TRUNC = 4, /** truncate file, for comp layout */
3308 LAYOUT_INTENT_RELEASE = 5, /** reserved for HSM release */
3309 LAYOUT_INTENT_RESTORE = 6, /** reserved for HSM restore */
3312 /* enqueue layout lock with intent */
3313 struct layout_intent {
3314 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3316 struct lu_extent li_extent;
3317 } __attribute__((packed));
3320 * On the wire version of hsm_progress structure.
3322 * Contains the userspace hsm_progress and some internal fields.
3324 struct hsm_progress_kernel {
3325 /* Field taken from struct hsm_progress */
3326 struct lu_fid hpk_fid;
3328 struct hsm_extent hpk_extent;
3330 __u16 hpk_errval; /* positive val */
3332 /* Additional fields */
3333 __u64 hpk_data_version;
3335 } __attribute__((packed));
3338 * OUT_UPDATE RPC Format
3340 * During the cross-ref operation, the Master MDT, which the client send the
3341 * request to, will disassembly the operation into object updates, then OSP
3342 * will send these updates to the remote MDT to be executed.
3344 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3345 * operation and does a type of modification to an object.
3353 * update (ub_count-th)
3355 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3360 * rc [+ buffers] (1st)
3361 * rc [+ buffers] (2st)
3363 * rc [+ buffers] (nr_count-th)
3365 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3366 * be equal to ub_count.
3370 * Type of each update, if adding/deleting update, please also update
3371 * update_opcode in lustre/target/out_lib.c.
3383 OUT_INDEX_LOOKUP = 9,
3384 OUT_INDEX_INSERT = 10,
3385 OUT_INDEX_DELETE = 11,
3391 OUT_XATTR_LIST = 17,
3396 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3397 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3398 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3399 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3402 struct object_update_param {
3403 __u16 oup_len; /* length of this parameter */
3410 struct object_update {
3411 __u16 ou_type; /* enum update_type */
3412 __u16 ou_params_count; /* update parameters count */
3413 __u32 ou_result_size; /* how many bytes can return */
3414 __u32 ou_flags; /* enum update_flag */
3415 __u32 ou_padding1; /* padding 1 */
3416 __u64 ou_batchid; /* op transno on master */
3417 struct lu_fid ou_fid; /* object to be updated */
3418 struct object_update_param ou_params[0]; /* update params */
3421 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3422 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3423 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3424 /* Hold object_updates sending to the remote OUT in single RPC */
3425 struct object_update_request {
3427 __u16 ourq_count; /* number of ourq_updates[] */
3429 struct object_update ourq_updates[0];
3432 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3433 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3434 /* Header for updates request between MDTs */
3435 struct out_update_header {
3438 __u32 ouh_inline_length;
3439 __u32 ouh_reply_size;
3440 __u32 ouh_inline_data[0];
3443 struct out_update_buffer {
3448 /* the result of object update */
3449 struct object_update_result {
3456 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3457 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3458 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3459 /* Hold object_update_results being replied from the remote OUT. */
3460 struct object_update_reply {
3467 /* read update result */
3468 struct out_read_reply {
3475 /** layout swap request structure
3476 * fid1 and fid2 are in mdt_body
3478 struct mdc_swap_layouts {
3480 } __attribute__((packed));
3482 #define INLINE_RESYNC_ARRAY_SIZE 15
3483 struct close_data_resync_done {
3485 __u32 resync_ids_inline[INLINE_RESYNC_ARRAY_SIZE];
3489 struct lustre_handle cd_handle;
3490 struct lu_fid cd_fid;
3491 __u64 cd_data_version;
3493 __u64 cd_reserved[8];
3494 struct close_data_resync_done cd_resync;
3498 __u32 cd_archive_id;
3502 /* Update llog format */
3504 struct lu_fid uop_fid;
3506 __u16 uop_param_count;
3507 __u16 uop_params_off[];
3508 } __attribute__((packed));
3511 struct update_op uops_op[0];
3514 struct update_params {
3515 struct object_update_param up_params[0];
3518 enum update_records_flag {
3519 UPDATE_RECORD_CONTINUE = 1 >> 0,
3522 * This is the update record format used to store the updates in
3523 * disk. All updates of the operation will be stored in ur_ops.
3524 * All of parameters for updates of the operation will be stored
3526 * To save the space of the record, parameters in ur_ops will only
3527 * remember their offset in ur_params, so to avoid storing duplicate
3528 * parameters in ur_params, which can help us save a lot space for
3529 * operation like creating striped directory.
3531 struct update_records {
3532 __u64 ur_master_transno;
3535 /* If the operation includes multiple updates, then ur_index
3536 * means the index of the update inside the whole updates. */
3538 __u32 ur_update_count;
3539 __u32 ur_param_count;
3540 struct update_ops ur_ops;
3541 /* Note ur_ops has a variable size, so comment out
3542 * the following ur_params, in case some use it directly
3543 * update_records->ur_params
3545 * struct update_params ur_params;
3549 struct llog_update_record {
3550 struct llog_rec_hdr lur_hdr;
3551 struct update_records lur_update_rec;
3552 /* Note ur_update_rec has a variable size, so comment out
3553 * the following ur_tail, in case someone use it directly
3555 * struct llog_rec_tail lur_tail;
3559 /* sepol string format is:
3560 * <1-digit for SELinux status>:<policy name>:<policy version>:<policy hash>
3562 /* Max length of the sepol string
3563 * Should be large enough to contain a sha512sum of the policy
3565 #define SELINUX_MODE_LEN 1
3566 #define SELINUX_POLICY_VER_LEN 3 /* 3 chars to leave room for the future */
3567 #define SELINUX_POLICY_HASH_LEN 64
3568 #define LUSTRE_NODEMAP_SEPOL_LENGTH (SELINUX_MODE_LEN + NAME_MAX + \
3569 SELINUX_POLICY_VER_LEN + \
3570 SELINUX_POLICY_HASH_LEN + 3)
3572 /* nodemap records, uses 32 byte record length */
3573 #define LUSTRE_NODEMAP_NAME_LENGTH 16
3574 struct nodemap_cluster_rec {
3575 char ncr_name[LUSTRE_NODEMAP_NAME_LENGTH + 1];
3579 __u32 ncr_squash_uid;
3580 __u32 ncr_squash_gid;
3583 /* lnet_nid_t is 8 bytes */
3584 struct nodemap_range_rec {
3585 lnet_nid_t nrr_start_nid;
3586 lnet_nid_t nrr_end_nid;
3591 struct nodemap_id_rec {
3599 struct nodemap_global_rec {
3610 struct nodemap_cluster_rec ncr;
3611 struct nodemap_range_rec nrr;
3612 struct nodemap_id_rec nir;
3613 struct nodemap_global_rec ngr;
3616 /* This is the lu_ladvise struct which goes out on the wire.
3617 * Corresponds to the userspace arg llapi_lu_ladvise.
3618 * value[1-4] are unspecified fields, used differently by different advices */
3620 __u16 lla_advice; /* advice type */
3621 __u16 lla_value1; /* values for different advice types */
3623 __u64 lla_start; /* first byte of extent for advice */
3624 __u64 lla_end; /* last byte of extent for advice */
3629 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3630 * arg llapi_ladvise_hdr.
3631 * value[1-3] are unspecified fields, used differently by different advices */
3632 struct ladvise_hdr {
3633 __u32 lah_magic; /* LADVISE_MAGIC */
3634 __u32 lah_count; /* number of advices */
3635 __u64 lah_flags; /* from enum ladvise_flag */
3636 __u32 lah_value1; /* unused */
3637 __u32 lah_value2; /* unused */
3638 __u64 lah_value3; /* unused */
3639 struct lu_ladvise lah_advise[0]; /* advices in this header */
3642 #if defined(__cplusplus)