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/nidstr.h>
74 #include <linux/lnet/lnet-types.h>
75 #include <linux/lustre/lustre_user.h>
76 #include <linux/lustre/lustre_ver.h>
78 #if defined(__cplusplus)
85 /* FOO_REQUEST_PORTAL is for incoming requests on the FOO
86 * FOO_REPLY_PORTAL is for incoming replies on the FOO
87 * FOO_BULK_PORTAL is for incoming bulk on the FOO
90 #define CONNMGR_REQUEST_PORTAL 1
91 #define CONNMGR_REPLY_PORTAL 2
92 #define OSC_REPLY_PORTAL 4
93 #define OST_IO_PORTAL 6
94 #define OST_CREATE_PORTAL 7
95 #define OST_BULK_PORTAL 8
96 #define MDC_REPLY_PORTAL 10
97 #define MDS_REQUEST_PORTAL 12
98 #define MDS_IO_PORTAL 13
99 #define MDS_BULK_PORTAL 14
100 #define LDLM_CB_REQUEST_PORTAL 15
101 #define LDLM_CB_REPLY_PORTAL 16
102 #define LDLM_CANCEL_REQUEST_PORTAL 17
103 #define LDLM_CANCEL_REPLY_PORTAL 18
104 /* #define MDS_SETATTR_PORTAL 22 obsolete after 2.13 */
105 #define MDS_READPAGE_PORTAL 23
106 #define OUT_PORTAL 24
107 #define MGC_REPLY_PORTAL 25
108 #define MGS_REQUEST_PORTAL 26
109 #define MGS_REPLY_PORTAL 27
110 #define OST_REQUEST_PORTAL 28
111 #define FLD_REQUEST_PORTAL 29
112 #define SEQ_METADATA_PORTAL 30
113 #define SEQ_DATA_PORTAL 31
114 #define SEQ_CONTROLLER_PORTAL 32
115 #define MGS_BULK_PORTAL 33
116 /* #define DVS_PORTAL 63 */
117 /* reserved for Cray DVS - spitzcor@cray.com, roe@cray.com, n8851@cray.com */
120 * Describes a range of sequence, lsr_start is included but lsr_end is
122 * Same structure is used in fld module where lsr_index field holds mdt id
125 struct lu_seq_range {
132 struct lu_seq_range_array {
135 struct lu_seq_range lsra_lsr[0];
138 #define LU_SEQ_RANGE_MDT 0x0
139 #define LU_SEQ_RANGE_OST 0x1
140 #define LU_SEQ_RANGE_ANY 0x3
142 #define LU_SEQ_RANGE_MASK 0x3
144 extern void lustre_lma_swab(struct lustre_mdt_attrs *lma);
145 extern void lustre_lma_init(struct lustre_mdt_attrs *lma,
146 const struct lu_fid *fid,
147 __u32 compat, __u32 incompat);
148 extern void lustre_loa_swab(struct lustre_ost_attrs *loa,
150 extern void lustre_loa_init(struct lustre_ost_attrs *loa,
151 const struct lu_fid *fid,
152 __u32 compat, __u32 incompat);
154 /* copytool can use any nonnegative integer to represent archive-Ids during
155 * register with MDT thru kuc.
156 * archive num = 0 => all
157 * archive num from 1 to MAX_U32
159 #define LL_HSM_ORIGIN_MAX_ARCHIVE (sizeof(__u32) * 8)
160 /* the max count of archive ids that one agent can support */
161 #define LL_HSM_MAX_ARCHIVES_PER_AGENT 1024
164 * HSM on-disk attributes stored in a separate xattr.
167 /** Bitfield for supported data in this structure. For future use. */
170 /** HSM flags, see hsm_flags enum below */
172 /** backend archive id associated with the file */
174 /** version associated with the last archiving, if any */
177 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
179 static inline void lov_foreign_hsm_to_cpu(struct lov_hsm_base *dst,
180 const struct lov_foreign_md *lfm)
182 struct lov_hsm_base *src = (struct lov_hsm_base *)lfm->lfm_value;
184 dst->lhb_archive_id = __le64_to_cpu(src->lhb_archive_id);
185 dst->lhb_archive_ver = __le64_to_cpu(src->lhb_archive_ver);
186 memcpy(dst->lhb_uuid, src->lhb_uuid, sizeof(dst->lhb_uuid));
189 static inline void lov_foreign_hsm_to_le(struct lov_foreign_md *lfm,
190 struct lov_hsm_base *src)
192 struct lov_hsm_base *dst = (struct lov_hsm_base *)lfm->lfm_value;
194 dst->lhb_archive_id = __cpu_to_le64(dst->lhb_archive_id);
195 dst->lhb_archive_ver = __cpu_to_le64(dst->lhb_archive_ver);
196 memcpy(dst->lhb_uuid, src->lhb_uuid, sizeof(dst->lhb_uuid));
203 /** LASTID file has zero OID */
204 LUSTRE_FID_LASTID_OID = 0UL,
205 /** initial fid id value */
206 LUSTRE_FID_INIT_OID = 1UL,
210 * Different FID Format
211 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
214 * File IDentifier generated by client from range allocated by the seq service.
215 * First 0x400 sequences [2^33, 2^33 + 0x400] are reserved for system use. Note
216 * that on ldiskfs MDTs that IGIF FIDs can use inode numbers starting at 12,
217 * but this is in the IGIF SEQ rangeand does not conflict with assigned FIDs.
220 * Inode and Generation In FID, a surrogate FID used to globally identify an
221 * existing object on OLD formatted MDT file system. This would only be used on
222 * MDT0 in a DNE filesystem, because there are not expected to be any OLD
223 * formatted DNE filesystems. Belongs to a sequence in [12, 2^32 - 1] range,
224 * where sequence number is inode number, and inode generation is used as OID.
225 * NOTE: This assumes no more than 2^32-1 inodes exist in the MDT filesystem,
226 * which is the maximum possible for an ldiskfs backend. NOTE: This assumes
227 * that the reserved ext3/ext4/ldiskfs inode numbers [0-11] are never visible
228 * to clients, which has always been true.
231 * Object ID in FID, a surrogate FID used to globally identify an existing
232 * object on OLD formatted OST file system. Belongs to a sequence in
233 * [2^32, 2^33 - 1]. Sequence number is calculated as:
234 * 1 << 32 | (ost_index << 16) | ((objid >> 32) & 0xffff)
235 * that is, SEQ consists of 16-bit OST index, and higher 16 bits of object ID.
236 * The generation of unique SEQ values per OST allows the IDIF FIDs to be
237 * identified in the FLD correctly. The OID field is calculated as:
239 * that is, it consists of lower 32 bits of object ID. NOTE This assumes that
240 * no more than 2^48-1 objects have ever been created on an OST, and that no
241 * more than 65535 OSTs are in use. Both are very reasonable assumptions (can
242 * uniquely map all objects on an OST that created 1M objects per second for 9
243 * years, or combinations thereof).
246 * Surrogate FID used to identify an existing object on OLD formatted OST
247 * filesystem. Belongs to the reserved sequence 0, and is used internally prior
248 * to the introduction of FID-on-OST, at which point IDIF will be used to
249 * identify objects as residing on a specific OST.
252 * For Lustre Log objects the object sequence 1 is used. This is compatible with
253 * both OLD and NEW.1 namespaces, as this SEQ number is in the ext3/ldiskfs
254 * reserved inode range and does not conflict with IGIF sequence numbers.
257 * For testing OST IO performance the object sequence 2 is used. This is
258 * compatible with both OLD and NEW.1 namespaces, as this SEQ number is in the
259 * ext3/ldiskfs reserved inode range and does not conflict with IGIF sequence
262 * OST_MDT1 .. OST_MAX:
263 * For testing with multiple MDTs the object sequence 3 through 9 is used,
264 * allowing direct mapping of MDTs 1 through 7 respectively, for a total of 8
265 * MDTs including OST_MDT0. This matches the legacy CMD project "group"
266 * mappings. However, this SEQ range is only for testing prior to any production
267 * DNE release, as the objects in this range conflict across all OSTs, as the
268 * OST index is not part of the FID.
271 * For compatibility with existing OLD OST network protocol structures, the FID
272 * must map onto the o_id and o_gr in a manner that ensures existing objects are
273 * identified consistently for IO, as well as onto the lock namespace to ensure
274 * both IDIFs map onto the same objects for IO as well as resources in the DLM.
277 * resource[] = {o_id, o_seq, 0, 0}; // o_seq == 0 for production releases
279 * DLM NEW.1 FID (this is the same for both the MDT and OST):
280 * resource[] = {SEQ, OID, VER, HASH};
282 * Note that for mapping IDIF values to DLM resource names the o_id may be
283 * larger than the 2^33 reserved sequence numbers for IDIF, so it is possible
284 * for the o_id numbers to overlap FID SEQ numbers in the resource. However, in
285 * all production releases the OLD o_seq field is always zero, and all valid FID
286 * OID values are non-zero, so the lock resources will not collide.
288 * For objects within the IDIF range, group extraction (non-CMD) will be:
289 * o_id = (fid->f_seq & 0x7fff) << 16 | fid->f_oid;
290 * o_seq = 0; // formerly group number
294 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
295 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
296 * used for other purposes and not risk collisions with existing inodes.
299 FID_SEQ_OST_MDT0 = 0,
300 FID_SEQ_LLOG = 1, /* unnamed llogs */
302 FID_SEQ_UNUSED_START = 3, /* Unused */
303 FID_SEQ_UNUSED_END = 9, /* Unused */
304 FID_SEQ_LLOG_NAME = 10, /* named llogs */
307 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
308 FID_SEQ_IDIF = 0x100000000ULL,
309 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
310 /* Normal FID sequence starts from this value, i.e. 1<<33 */
311 FID_SEQ_START = 0x200000000ULL,
312 /* sequence for local pre-defined FIDs listed in local_oid */
313 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
314 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
315 /* named FIDs generated by local_object_storage library */
316 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
317 /* current FLD will only cache the fid seq, instead of oid on the client
318 * side, if the FID needs to be exposed to clients, it needs to make
319 * sure all of fids under one seq will be located in one MDT.
321 FID_SEQ_SPECIAL = 0x200000004ULL,
322 FID_SEQ_QUOTA = 0x200000005ULL,
323 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
324 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
325 FID_SEQ_LAYOUT_RBTREE = 0x200000008ULL,
326 /* sequence is used for update logs of cross-MDT operation */
327 FID_SEQ_UPDATE_LOG = 0x200000009ULL,
328 /* DNE recovery logs in update_logs_dir */
329 FID_SEQ_UPDATE_LOG_DIR = 0x20000000aULL,
330 FID_SEQ_NORMAL = 0x200000400ULL,
331 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
334 #define OBIF_OID_MAX_BITS 32
335 #define OBIF_MAX_OID ((1ULL << OBIF_OID_MAX_BITS) - 1)
336 #define IDIF_OID_MAX_BITS 48
337 #define IDIF_MAX_OID ((1ULL << IDIF_OID_MAX_BITS) - 1)
339 /** OID for FID_SEQ_SPECIAL */
341 /* Big Filesystem Lock to serialize rename operations */
342 FID_OID_SPECIAL_BFL = 1UL,
345 /** OID for FID_SEQ_DOT_LUSTRE */
346 enum dot_lustre_oid {
347 FID_OID_DOT_LUSTRE = 1UL,
348 FID_OID_DOT_LUSTRE_OBF = 2UL,
349 FID_OID_DOT_LUSTRE_LPF = 3UL,
352 /** OID for FID_SEQ_ROOT */
355 FID_OID_ECHO_ROOT = 2UL,
358 struct lu_orphan_rec {
359 /* The MDT-object's FID referenced by the orphan OST-object */
360 struct lu_fid lor_fid;
365 struct lu_orphan_ent {
366 /* The orphan OST-object's FID */
367 struct lu_fid loe_key;
368 struct lu_orphan_rec loe_rec;
371 struct lu_orphan_rec_v2 {
372 struct lu_orphan_rec lor_rec;
373 struct ost_layout lor_layout;
377 struct lu_orphan_ent_v2 {
378 /* The orphan OST-object's FID */
379 struct lu_fid loe_key;
380 struct lu_orphan_rec_v2 loe_rec;
383 struct lu_orphan_rec_v3 {
384 struct lu_orphan_rec lor_rec;
385 struct ost_layout lor_layout;
386 /* The OST-object declared layout version in PFID EA.*/
387 __u32 lor_layout_version;
388 /* The OST-object declared layout range (of version) in PFID EA.*/
394 struct lu_orphan_ent_v3 {
395 /* The orphan OST-object's FID */
396 struct lu_fid loe_key;
397 struct lu_orphan_rec_v3 loe_rec;
402 /** \defgroup lu_dir lu_dir
407 * Enumeration of possible directory entry attributes.
409 * Attributes follow directory entry header in the order they appear in this
412 enum lu_dirent_attrs {
415 LUDA_64BITHASH = 0x0004,
417 /* for MDT internal use only, not visible to client */
419 /* Something in the record is unknown, to be verified in further. */
420 LUDA_UNKNOWN = 0x0400,
421 /* Ignore this record, go to next directly. */
422 LUDA_IGNORE = 0x0800,
423 /* The system is upgraded, has beed or to be repaired (dryrun). */
424 LUDA_UPGRADE = 0x1000,
425 /* The dirent has been repaired, or to be repaired (dryrun). */
426 LUDA_REPAIR = 0x2000,
427 /* Only check but not repair the dirent inconsistency */
428 LUDA_VERIFY_DRYRUN = 0x4000,
429 /* Verify the dirent consistency */
430 LUDA_VERIFY = 0x8000,
433 #define LU_DIRENT_ATTRS_MASK 0xff00
436 * Layout of readdir pages, as transmitted on wire.
439 /** valid if LUDA_FID is set. */
440 struct lu_fid lde_fid;
441 /** a unique entry identifier: a hash or an offset. */
443 /** total record length, including all attributes. */
447 /** optional variable size attributes following this entry.
448 * taken from enum lu_dirent_attrs.
451 /** name is followed by the attributes indicated in ->ldp_attrs, in
452 * their natural order. After the last attribute, padding bytes are
453 * added to make ->lde_reclen a multiple of 8.
459 * Definitions of optional directory entry attributes formats.
461 * Individual attributes do not have their length encoded in a generic way. It
462 * is assumed that consumer of an attribute knows its format. This means that
463 * it is impossible to skip over an unknown attribute, except by skipping over
464 * all remaining attributes (by using ->lde_reclen), which is not too
465 * constraining, because new server versions will append new attributes at
466 * the end of an entry.
470 * Fid directory attribute: a fid of an object referenced by the entry. This
471 * will be almost always requested by the client and supplied by the server.
473 * Aligned to 8 bytes.
475 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
480 * Aligned to 2 bytes.
487 __u64 ldp_hash_start;
491 struct lu_dirent ldp_entries[0];
494 enum lu_dirpage_flags {
495 LDF_EMPTY = 1 << 0, /* dirpage contains no entry. */
496 LDF_COLLIDE = 1 << 1, /* last entry lde_hash == ldp_hash_end */
499 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
501 if (__le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
504 return dp->ldp_entries;
507 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
509 struct lu_dirent *next;
511 if (__le16_to_cpu(ent->lde_reclen) != 0)
512 next = ((void *)ent) + __le16_to_cpu(ent->lde_reclen);
519 static inline __kernel_size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
521 __kernel_size_t size;
523 if (attr & LUDA_TYPE) {
524 const __kernel_size_t align = sizeof(struct luda_type) - 1;
526 size = (sizeof(struct lu_dirent) + namelen + 1 + align) &
528 size += sizeof(struct luda_type);
530 size = sizeof(struct lu_dirent) + namelen + 1;
533 return (size + 7) & ~7;
536 static inline __u16 lu_dirent_type_get(struct lu_dirent *ent)
539 struct luda_type *lt;
542 if (__le32_to_cpu(ent->lde_attrs) & LUDA_TYPE) {
543 const unsigned int align = sizeof(struct luda_type) - 1;
545 len = __le16_to_cpu(ent->lde_namelen);
546 len = (len + align) & ~align;
547 lt = (void *)ent->lde_name + len;
548 type = __le16_to_cpu(lt->lt_type);
554 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
557 * MDS_READPAGE page size
559 * This is the directory page size packed in MDS_READPAGE RPC.
560 * It's different than PAGE_SIZE because the client needs to
561 * access the struct lu_dirpage header packed at the beginning of
562 * the "page" and without this there isn't any way to know find the
563 * lu_dirpage header is if client and server PAGE_SIZE differ.
565 #define LU_PAGE_SHIFT 12
566 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
567 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
569 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
573 struct lustre_handle {
576 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
578 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
580 return lh->cookie != 0;
583 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
584 const struct lustre_handle *lh2)
586 return lh1->cookie == lh2->cookie;
589 static inline void lustre_handle_copy(struct lustre_handle *tgt,
590 const struct lustre_handle *src)
592 tgt->cookie = src->cookie;
595 /* lustre_msg struct magic. DON'T use swabbed values of MAGIC as magic! */
596 enum lustre_msg_magic {
597 LUSTRE_MSG_MAGIC_V2 = 0x0BD00BD3,
598 LUSTRE_MSG_MAGIC_V2_SWABBED = 0xD30BD00B,
599 LUSTRE_MSG_MAGIC = LUSTRE_MSG_MAGIC_V2
602 /* flags for lm_flags */
604 MSGHDR_AT_SUPPORT = 0x1, /* lm_cksum valid early reply msg */
605 MSGHDR_CKSUM_INCOMPAT18 = 0x2, /* compat for 1.8, need beyond 2.8.0 */
608 #define lustre_msg lustre_msg_v2
609 /* we depend on this structure to be 8-byte aligned */
610 /* this type is only endian-adjusted in lustre_unpack_msg() */
611 struct lustre_msg_v2 {
612 __u32 lm_bufcount; /* number of buffers in lm_buflens[] */
613 __u32 lm_secflvr; /* 0 = no crypto, or sptlrpc security flavour */
614 __u32 lm_magic; /* RPC version magic = LUSTRE_MSG_MAGIC_V2 */
615 __u32 lm_repsize; /* size of preallocated reply buffer */
616 __u32 lm_cksum; /* CRC32 of ptlrpc_body early reply messages */
617 __u32 lm_flags; /* enum lustre_msghdr MSGHDR_* flags */
618 __u32 lm_opc; /* SUB request opcode in a batch request */
619 __u32 lm_padding_3; /* unused */
620 /* length of additional buffers in bytes
621 * message buffers are packed after padded lm_buflens[] array,
622 * padded to a multiple of 8 bytes each to align contents.
627 /* The returned result of the SUB request in a batch request */
628 #define lm_result lm_opc
630 /* ptlrpc_body packet pb_types */
631 #define PTL_RPC_MSG_REQUEST 4711 /* normal RPC request message */
632 #define PTL_RPC_MSG_ERR 4712 /* error reply if request unprocessed */
633 #define PTL_RPC_MSG_REPLY 4713 /* normal RPC reply message */
635 /* ptlrpc_body pb_version ((target_version << 16) | rpc_version) */
636 enum lustre_msg_version {
637 PTLRPC_MSG_VERSION = 0x00000003,
638 LUSTRE_VERSION_MASK = 0xffff0000,
639 LUSTRE_OBD_VERSION = 0x00010000,
640 LUSTRE_MDS_VERSION = 0x00020000,
641 LUSTRE_OST_VERSION = 0x00030000,
642 LUSTRE_DLM_VERSION = 0x00040000,
643 LUSTRE_LOG_VERSION = 0x00050000,
644 LUSTRE_MGS_VERSION = 0x00060000,
647 /* pb_flags that apply to all request messages */
648 /* #define MSG_LAST_REPLAY 0x0001 obsolete 2.0 => {REQ,LOCK}_REPLAY_DONE */
649 #define MSG_RESENT 0x0002 /* was previously sent, no reply seen */
650 #define MSG_REPLAY 0x0004 /* was processed, got reply, recovery */
651 /* #define MSG_AT_SUPPORT 0x0008 obsolete since 1.5, AT always enabled */
652 /* #define MSG_DELAY_REPLAY 0x0010 obsolete since 2.0 */
653 /* #define MSG_VERSION_REPLAY 0x0020 obsolete since 1.8.2, VBR always on */
654 #define MSG_REQ_REPLAY_DONE 0x0040 /* request replay over, locks next */
655 #define MSG_LOCK_REPLAY_DONE 0x0080 /* lock replay over, client done */
657 /* pb_op_flags for connect opcodes: MDS_CONNECT, OST_CONNECT, MGS_CONNECT */
658 #define MSG_CONNECT_RECOVERING 0x00000001 /* target is in recovery */
659 #define MSG_CONNECT_RECONNECT 0x00000002 /* tgt already has client import */
660 #define MSG_CONNECT_REPLAYABLE 0x00000004 /* target supports RPC replay */
661 /* #define MSG_CONNECT_PEER 0x00000008 obsolete since 1.2, removed in 1.5 */
662 #define MSG_CONNECT_LIBCLIENT 0x00000010 /* obsolete since 2.3, removed 2.6 */
663 #define MSG_CONNECT_INITIAL 0x00000020 /* first client connection attempt */
664 /* #define MSG_CONNECT_ASYNC 0x00000040 obsolete since 1.5 */
665 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
666 #define MSG_CONNECT_TRANSNO 0x00000100 /* client sent transno in replay */
667 #define MSG_PACK_UID_GID 0x00000200 /* thread UID/GID in ptlrpc_body */
669 /* number of previous object versions in pb_pre_versions[] */
670 #define PTLRPC_NUM_VERSIONS 4
671 /* without gss, ptlrpc_body is put at the first buffer. */
672 struct ptlrpc_body_v3 {
673 struct lustre_handle pb_handle;
674 __u32 pb_type; /* request/reply/err type: PTL_RPC_MSG_* */
675 __u32 pb_version; /* LUSTRE_*_VERSION | PTLRPC_MSG_VERSION */
676 __u32 pb_opc; /* RPC opcodes: MDS_*, OST_*, LDLM_, ... */
677 __u32 pb_status; /* negative Linux x86 error number */
678 __u64 pb_last_xid; /* highest replied XID w/o lower unreplied XID*/
679 __u16 pb_tag; /* multiple modifying RPCs virtual slot index */
682 __u64 pb_last_committed;/* rep: highest pb_transno committed to disk */
683 __u64 pb_transno; /* server-assigned transno for modifying RPCs */
684 __u32 pb_flags; /* req: MSG_* flags */
685 __u32 pb_op_flags; /* req: MSG_CONNECT_* flags */
686 __u32 pb_conn_cnt; /* connect instance of this client on server */
687 __u32 pb_timeout; /* req: max wait time; rep: service estimate */
688 __u32 pb_service_time; /* rep: server arrival to reply in seconds */
689 __u32 pb_limit; /* rep: dynamic DLM LRU lock count limit */
690 __u64 pb_slv; /* rep: dynamic DLM LRU server lock volume */
691 /* VBR: rep: previous pb_version(s) of objects modified by this RPC */
692 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
693 __u64 pb_mbits; /**< match bits for bulk request */
694 /* padding for future needs - fix lustre_swab_ptlrpc_body() also */
695 __u64 pb_padding64_0;
696 __u64 pb_padding64_1;
697 __u32 pb_uid; /* req: process uid, use by tbf rules */
698 __u32 pb_gid; /* req: process gid, use by tbf rules */
699 char pb_jobid[LUSTRE_JOBID_SIZE]; /* req: ASCII jobid from env + NUL */
701 #define ptlrpc_body ptlrpc_body_v3
703 struct ptlrpc_body_v2 {
704 struct lustre_handle pb_handle;
709 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
710 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
713 __u64 pb_last_committed;
718 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
719 __u32 pb_service_time; /* rep: actual service time, req: net_latency */
722 /* VBR: pre-versions */
723 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
724 __u64 pb_mbits; /**< unused in V2 */
725 /* padding for future needs */
726 __u64 pb_padding64_0;
727 __u64 pb_padding64_1;
728 __u32 pb_uid; /* req: process uid, use by tbf rules */
729 __u32 pb_gid; /* req: process gid, use by tbf rules */
732 /* message body offset for lustre_msg_v2 */
733 /* ptlrpc body offset in all request/reply messages */
734 #define MSG_PTLRPC_BODY_OFF 0
736 /* normal request/reply message record offset */
737 #define REQ_REC_OFF 1
738 #define REPLY_REC_OFF 1
740 /* ldlm request message body offset */
741 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
742 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
744 /* ldlm intent lock message body offset */
745 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
746 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
748 /* ldlm reply message body offset */
749 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
750 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
752 /** only use in req->rq_{req,rep}_swab_mask */
753 #define MSG_PTLRPC_HEADER_OFF 31
756 #define OBD_CONNECT_RDONLY 0x1ULL /* client is read-only */
757 #define OBD_CONNECT_INDEX 0x2ULL /* connect to LOV idx */
758 #define OBD_CONNECT_MDS 0x4ULL /* connect MDT to OST */
759 #define OBD_CONNECT_GRANT 0x8ULL /* fetch grant connect */
760 #define OBD_CONNECT_SRVLOCK 0x10ULL /* server lock for RPC */
761 #define OBD_CONNECT_VERSION 0x20ULL /* versions in OCD */
762 #define OBD_CONNECT_REQPORTAL 0x40ULL /* non-IO portal */
763 #define OBD_CONNECT_ACL 0x80ULL /* access control list */
764 #define OBD_CONNECT_XATTR 0x100ULL /* extended attributes */
765 #define OBD_CONNECT_LARGE_ACL 0x200ULL /* over 32 ACL entries */
766 /* was OBD_CONNECT_TRUNCLOCK 0x400ULL * server locks punch */
767 /* temporary reuse until 2.21.53 to indicate pre-2.15 client, see LU-15478 */
768 #define OBD_CONNECT_OLD_FALLOC 0x400ULL /* no o_valid flags */
769 #define OBD_CONNECT_TRANSNO 0x800ULL /* replay send transno */
770 #define OBD_CONNECT_IBITS 0x1000ULL /* not checked 2.11+ */
771 #define OBD_CONNECT_BARRIER 0x2000ULL /* write barrier */
772 #define OBD_CONNECT_ATTRFID 0x4000ULL /* Server GetAttr FID */
773 #define OBD_CONNECT_NODEVOH 0x8000ULL /* No open handle spec */
774 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Never used, gone 2.9*/
775 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Never used, gone 2.9*/
776 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /* Max bytes per rpc */
777 #define OBD_CONNECT_QUOTA64 0x80000ULL /* Unused since 2.4 */
778 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /* Unused since 2.7 */
779 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /* Unused since 2.7 */
780 #define OBD_CONNECT_CANCELSET 0x400000ULL /* Early batch cancel */
781 #define OBD_CONNECT_SOM 0x800000ULL /* Unused since 2.7 */
782 #define OBD_CONNECT_AT 0x1000000ULL /* client uses AT */
783 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /* LRU resize feature */
784 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /* MDS-MDS connection */
785 #define OBD_CONNECT_REAL 0x8000000ULL /* Unused since 2.8 */
786 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /* Unused since 2.4 */
787 #define OBD_CONNECT_CKSUM 0x20000000ULL /* cksum algo choice */
788 #define OBD_CONNECT_FID 0x40000000ULL /* server handles FIDs */
789 #define OBD_CONNECT_VBR 0x80000000ULL /* version based recov */
790 #define OBD_CONNECT_LOV_V3 0x100000000ULL /* client LOV v3 EA */
791 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* handle grant shrink */
792 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* no orph objid reuse */
793 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* EA size in reply */
794 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
795 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client layout lock */
796 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* 64-bits dir hash */
797 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
798 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recov support */
799 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* ptlrpc_body jobid */
800 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create client umask */
801 /* client -EINPROGRESS RPC error handling */
802 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL
803 /* extra grant params for space reservation */
804 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL
805 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* unused since 2.0 */
806 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable LVB type */
807 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosec timestamp */
808 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL /* lightweight connect */
809 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL /* short io */
810 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL /* pings not required */
811 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL /* flk deadlock detect */
812 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL /* create stripe disp */
813 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by FID won't pack
816 #define OBD_CONNECT_LFSCK 0x40000000000000ULL /* allow online LFSCK */
817 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL /* unlink closes file */
818 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* allow multiple change
821 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
822 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
823 /* was OBD_CONNECT_LOCKAHEAD_OLD 0x1000000000000000ULL old lockahead 2.12-2.13*/
824 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL /* ptlrpc_body matchbit*/
825 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
826 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
827 /* ocd_connect_flags2 flags */
828 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* security context */
829 #define OBD_CONNECT2_LOCKAHEAD 0x2ULL /* ladvise lockahead */
830 #define OBD_CONNECT2_DIR_MIGRATE 0x4ULL /* migrate striped dir */
831 #define OBD_CONNECT2_SUM_STATFS 0x8ULL /* MDT aggregate statfs*/
832 #define OBD_CONNECT2_OVERSTRIPING 0x10ULL /* OST overstriping */
833 #define OBD_CONNECT2_FLR 0x20ULL /* FLR mirror handling */
834 #define OBD_CONNECT2_WBC_INTENTS 0x40ULL /* MDS wb cache intent */
835 #define OBD_CONNECT2_LOCK_CONVERT 0x80ULL /* IBITS lock convert */
836 #define OBD_CONNECT2_ARCHIVE_ID_ARRAY 0x100ULL /* HSM archive_id array*/
837 #define OBD_CONNECT2_INC_XID 0x200ULL /* Increasing xid */
838 #define OBD_CONNECT2_SELINUX_POLICY 0x400ULL /* cli SELinux policy */
839 #define OBD_CONNECT2_LSOM 0x800ULL /* Lazy Size on MDT */
840 #define OBD_CONNECT2_PCC 0x1000ULL /* Persist Client Cache*/
841 #define OBD_CONNECT2_CRUSH 0x2000ULL /* CRUSH dir hash */
842 #define OBD_CONNECT2_ASYNC_DISCARD 0x4000ULL /* async DoM discard */
843 #define OBD_CONNECT2_ENCRYPT 0x8000ULL /* client disk encrypt */
844 #define OBD_CONNECT2_FIDMAP 0x10000ULL /* MDT migrate FID map */
845 #define OBD_CONNECT2_GETATTR_PFID 0x20000ULL /* parent FID getattr */
846 #define OBD_CONNECT2_LSEEK 0x40000ULL /* SEEK_HOLE/DATA RPC */
847 #define OBD_CONNECT2_DOM_LVB 0x80000ULL /* DoM glimpse in LVB */
848 #define OBD_CONNECT2_REP_MBITS 0x100000ULL /* reply mbits, not XID*/
849 #define OBD_CONNECT2_MODE_CONVERT 0x200000ULL /* LDLM mode convert */
850 #define OBD_CONNECT2_BATCH_RPC 0x400000ULL /* Multi-op batch RPCs */
851 #define OBD_CONNECT2_PCCRO 0x800000ULL /* PCC read-only */
852 #define OBD_CONNECT2_MNE_TYPE 0x1000000ULL /* mne_nid_type IPv6 */
853 #define OBD_CONNECT2_LOCK_CONTENTION 0x2000000ULL /* contention detect */
854 #define OBD_CONNECT2_ATOMIC_OPEN_LOCK 0x4000000ULL /* lock on first open */
855 #define OBD_CONNECT2_ENCRYPT_NAME 0x8000000ULL /* name encrypt */
856 /* client handle DMV inheritance */
857 #define OBD_CONNECT2_DMV_IMP_INHERIT 0x20000000ULL
858 #define OBD_CONNECT2_ENCRYPT_FID2PATH 0x40000000ULL /* fid2path enc file */
859 /* For MDS+OSS rolling upgrade interop with 2.16+older, ignored after 2.20.53 */
860 #define OBD_CONNECT2_REPLAY_CREATE 0x80000000ULL /* replay OST_CREATE */
861 /* understands large/IPv6 NIDs */
862 #define OBD_CONNECT2_LARGE_NID 0x100000000ULL
863 #define OBD_CONNECT2_COMPRESS 0x200000000ULL /* compressed file */
864 /* only ZFS servers require a change to support unaligned DIO, so this flag is
865 * ignored for ldiskfs servers
867 #define OBD_CONNECT2_UNALIGNED_DIO 0x400000000ULL /* unaligned DIO */
868 #define OBD_CONNECT2_CONN_POLICY 0x800000000ULL /* server-side connection policy */
869 /* XXX README XXX README XXX README XXX README XXX README XXX README XXX
870 * Please DO NOT add OBD_CONNECT flags before first ensuring that this value
871 * is not in use by some other branch/patch. Email adilger@whamcloud.com
872 * to reserve the new OBD_CONNECT value for use by your feature. Then, submit
873 * a small patch against master and LTS branches that ONLY adds the new flag,
874 * updates obd_connect_names[], adds the flag to check_obd_connect_data(),
875 * and updates wiretests accordingly, so it can be approved and landed easily
876 * to reserve the flag for future use by your feature (submitted separately).
879 #define OCD_HAS_FLAG(ocd, flag) \
880 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flag))
881 #define OCD_HAS_FLAG2(ocd, flag2) (OCD_HAS_FLAG(ocd, FLAGS2) && \
882 !!((ocd)->ocd_connect_flags2 & OBD_CONNECT2_##flag2))
885 #ifdef HAVE_LRU_RESIZE_SUPPORT
886 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
888 #define LRU_RESIZE_CONNECT_FLAG 0
891 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
892 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
893 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
894 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
895 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
896 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
897 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
898 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
899 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
900 OBD_CONNECT_EINPROGRESS | \
901 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
902 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
903 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
904 OBD_CONNECT_FLOCK_DEAD | \
905 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
906 OBD_CONNECT_OPEN_BY_FID | \
907 OBD_CONNECT_DIR_STRIPE | OBD_CONNECT_GRANT | \
908 OBD_CONNECT_SRVLOCK | OBD_CONNECT_BULK_MBITS |\
910 OBD_CONNECT_MULTIMODRPCS |\
911 OBD_CONNECT_SUBTREE | OBD_CONNECT_LARGE_ACL |\
912 OBD_CONNECT_GRANT_PARAM | \
913 OBD_CONNECT_GRANT_SHRINK | \
914 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
916 #define MDT_CONNECT_SUPPORTED2 (OBD_CONNECT2_FILE_SECCTX | \
917 OBD_CONNECT2_DIR_MIGRATE | \
918 OBD_CONNECT2_SUM_STATFS | \
919 OBD_CONNECT2_OVERSTRIPING | \
921 OBD_CONNECT2_LOCK_CONVERT | \
922 OBD_CONNECT2_ARCHIVE_ID_ARRAY | \
923 OBD_CONNECT2_INC_XID | \
924 OBD_CONNECT2_SELINUX_POLICY | \
925 OBD_CONNECT2_LSOM | \
926 OBD_CONNECT2_ASYNC_DISCARD | \
928 OBD_CONNECT2_CRUSH | \
929 OBD_CONNECT2_ENCRYPT | \
930 OBD_CONNECT2_GETATTR_PFID |\
931 OBD_CONNECT2_LSEEK | OBD_CONNECT2_DOM_LVB |\
932 OBD_CONNECT2_REP_MBITS | \
933 OBD_CONNECT2_ATOMIC_OPEN_LOCK | \
934 OBD_CONNECT2_BATCH_RPC | \
935 OBD_CONNECT2_ENCRYPT_NAME | \
936 OBD_CONNECT2_ENCRYPT_FID2PATH | \
937 OBD_CONNECT2_DMV_IMP_INHERIT |\
938 OBD_CONNECT2_UNALIGNED_DIO)
940 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
941 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
942 OBD_CONNECT_INDEX | \
943 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
944 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
945 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
946 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
947 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
948 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
949 OBD_CONNECT_MAX_EASIZE | \
950 OBD_CONNECT_EINPROGRESS | \
951 OBD_CONNECT_JOBSTATS | \
952 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
953 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
954 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
955 OBD_CONNECT_BULK_MBITS | \
956 OBD_CONNECT_GRANT_PARAM | \
957 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
959 #define OST_CONNECT_SUPPORTED2 (OBD_CONNECT2_LOCKAHEAD | OBD_CONNECT2_INC_XID |\
960 OBD_CONNECT2_ENCRYPT | OBD_CONNECT2_LSEEK |\
961 OBD_CONNECT2_REP_MBITS |\
962 OBD_CONNECT2_REPLAY_CREATE |\
963 OBD_CONNECT2_UNALIGNED_DIO)
965 #define ECHO_CONNECT_SUPPORTED (OBD_CONNECT_FID | OBD_CONNECT_FLAGS2)
966 #define ECHO_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS
968 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
969 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
970 OBD_CONNECT_PINGLESS |\
971 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_BARRIER | \
974 #define MGS_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS | \
975 OBD_CONNECT2_LARGE_NID
977 /* Features required for this version of the client to work with server */
978 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_FID | \
979 OBD_CONNECT_ATTRFID | \
982 /* This structure is used for both request and reply.
984 * If we eventually have separate connect data for different types, which we
985 * almost certainly will, then perhaps we stick a union in here.
987 struct obd_connect_data {
988 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
989 __u32 ocd_version; /* lustre release version number */
990 __u32 ocd_grant; /* initial cache grant amount (bytes) */
991 __u32 ocd_index; /* LOV index to connect to */
992 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
993 __u64 ocd_ibits_known; /* inode bits this client understands */
994 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
995 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
996 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
997 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
998 __u64 ocd_transno; /* first transno from client to be replayed */
999 __u32 ocd_group; /* MDS group on OST */
1000 __u32 ocd_cksum_types; /* supported checksum algorithms */
1001 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
1002 __u32 ocd_instance; /* instance # of this target */
1003 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
1004 /* Fields after ocd_maxbytes are only accessible by the receiver
1005 * if the corresponding flag in ocd_connect_flags is set. Accessing
1006 * any field after ocd_maxbytes on the receiver without a valid flag
1007 * may result in out-of-bound memory access and kernel oops.
1009 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
1010 __u8 ocd_conn_policy; /* bitmask for client/server to communicate about policy */
1011 __u8 padding0; /* READ BELOW! also fix lustre_swab_connect */
1012 __u32 padding1; /* READ BELOW! also fix lustre_swab_connect */
1013 __u64 ocd_connect_flags2;/* OBD_CONNECT2_* per above */
1014 __u64 ocd_compr_type; /* bitmask of supported compression types */
1015 __u64 padding4; /* READ BELOW! also fix lustre_swab_connect */
1016 __u64 padding5; /* READ BELOW! also fix lustre_swab_connect */
1017 __u64 padding6; /* READ BELOW! also fix lustre_swab_connect */
1018 __u64 padding7; /* READ BELOW! also fix lustre_swab_connect */
1019 __u64 padding8; /* READ BELOW! also fix lustre_swab_connect */
1020 __u64 padding9; /* READ BELOW! also fix lustre_swab_connect */
1021 __u64 paddingA; /* READ BELOW! also fix lustre_swab_connect */
1022 __u64 paddingB; /* READ BELOW! also fix lustre_swab_connect */
1023 __u64 paddingC; /* READ BELOW! also fix lustre_swab_connect */
1024 __u64 paddingD; /* READ BELOW! also fix lustre_swab_connect */
1025 __u64 paddingE; /* READ BELOW! also fix lustre_swab_connect */
1026 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1028 /* XXX README XXX README XXX README XXX README XXX README XXX README XXX
1029 * Please DO NOT use any fields before first ensuring that this field is
1030 * not in use by some other branch/patch. Email adilger@whamcloud.com to
1031 * reserve the new obd_connect_data field for use by your feature. Then, submit
1032 * a small patch against master and LTS branch that ONLY adds the new field,
1033 * updates lustre_swab_connect(), along with the matching OBD_CONNECT flag,
1034 * and updates wiretests accordingly,so it can be approved and landed easily
1035 * to reserve the field for future use by your feature (submitted separately).
1039 * Supported checksum algorithms. Up to 32 checksum types are supported.
1040 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1041 * Please update DECLARE_CKSUM_NAME in obd_cksum.h when adding a new
1042 * algorithm and also the OBD_FL_CKSUM* flags, OBD_CKSUM_ALL flag,
1043 * OBD_FL_CKSUM_ALL flag and potentially OBD_CKSUM_T10_ALL flag.
1046 OBD_CKSUM_CRC32 = 0x00000001,
1047 OBD_CKSUM_ADLER = 0x00000002,
1048 OBD_CKSUM_CRC32C = 0x00000004,
1049 OBD_CKSUM_RESERVED = 0x00000008,
1050 OBD_CKSUM_T10IP512 = 0x00000010,
1051 OBD_CKSUM_T10IP4K = 0x00000020,
1052 OBD_CKSUM_T10CRC512 = 0x00000040,
1053 OBD_CKSUM_T10CRC4K = 0x00000080,
1056 #define OBD_CKSUM_T10_ALL (OBD_CKSUM_T10IP512 | OBD_CKSUM_T10IP4K | \
1057 OBD_CKSUM_T10CRC512 | OBD_CKSUM_T10CRC4K)
1059 #define OBD_CKSUM_ALL (OBD_CKSUM_CRC32 | OBD_CKSUM_ADLER | OBD_CKSUM_CRC32C | \
1063 * The default checksum algorithm used on top of T10PI GRD tags for RPC.
1064 * Considering that the checksum-of-checksums is only computing CRC32 on a
1065 * 4KB chunk of GRD tags for a 1MB RPC for 512B sectors, or 16KB of GRD
1066 * tags for 16MB of 4KB sectors, this is only 1/256 or 1/1024 of the
1067 * total data being checksummed, so the checksum type used here should not
1068 * affect overall system performance noticeably.
1070 #define OBD_CKSUM_T10_TOP OBD_CKSUM_ADLER
1073 * Explicitly define the meaning of the bits in ocd_conn_policy,
1074 * for server-side connection policy.
1076 enum conn_policy_flags {
1077 BYPASS_SOFT_BLOCK = 0,
1078 CONN_POLICY_FLAGS_NUM = 8,
1082 * OST requests: OBDO & OBD request records
1087 OST_REPLY = 0, /* reply ? */
1103 OST_QUOTACHECK = 18, /* not used since 2.4 */
1105 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1109 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
1111 #define OST_FIRST_OPC OST_REPLY
1114 OBD_FL_INLINEDATA = 0x00000001,
1115 OBD_FL_OBDMDEXISTS = 0x00000002,
1116 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1117 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1118 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1119 OBD_FL_RECREATE_OBJS = 0x00000020, /* recreate missing obj */
1120 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1121 OBD_FL_NO_PRJQUOTA = 0x00000080, /* the object's project is over
1124 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1125 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1126 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1127 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1128 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1129 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1130 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1131 OBD_FL_CKSUM_T10IP512 = 0x00005000, /* T10PI IP cksum, 512B sector */
1132 OBD_FL_CKSUM_T10IP4K = 0x00006000, /* T10PI IP cksum, 4KB sector */
1133 OBD_FL_CKSUM_T10CRC512 = 0x00007000, /* T10PI CRC cksum, 512B sector */
1134 OBD_FL_CKSUM_T10CRC4K = 0x00008000, /* T10PI CRC cksum, 4KB sector */
1135 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1136 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1137 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1138 * XXX: obsoleted - reserved for old
1139 * clients prior than 2.2
1141 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1142 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1143 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1144 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1145 OBD_FL_ROOT_SQUASH = 0x00800000, /* root squash */
1146 OBD_FL_ROOT_PRJQUOTA = 0x01000000, /* check prj quota for root */
1147 /* OBD_FL_LOCAL_MASK = 0xF0000000, was local-only flags until 2.10 */
1150 * Note that while the original checksum values were separate bits,
1151 * in 2.x we can actually allow all values from 1-31. T10-PI checksum
1152 * types already use values which are not separate bits.
1154 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1155 OBD_FL_CKSUM_CRC32C | OBD_FL_CKSUM_T10IP512 |
1156 OBD_FL_CKSUM_T10IP4K | OBD_FL_CKSUM_T10CRC512 |
1157 OBD_FL_CKSUM_T10CRC4K,
1159 OBD_FL_NO_QUOTA_ALL = OBD_FL_NO_USRQUOTA | OBD_FL_NO_GRPQUOTA |
1164 * All LOV EA magics should have the same postfix, if some new version
1165 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1166 * Lustre, even though the old version system does not recognizes such
1167 * new magic, it still can distinguish the corrupted cases by checking
1168 * the magic's postfix.
1170 #define LOV_MAGIC_MAGIC 0x0BD0
1171 #define LOV_MAGIC_MASK 0xFFFF
1173 #define LOV_MAGIC_BAD (0x0BAD0000 | LOV_MAGIC_MAGIC)
1174 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1175 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1176 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1177 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1178 /* reserved for specifying OSTs */
1179 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1180 #define LOV_MAGIC LOV_MAGIC_V1
1181 #define LOV_MAGIC_COMP_V1 (0x0BD60000 | LOV_MAGIC_MAGIC)
1182 #define LOV_MAGIC_FOREIGN (0x0BD70000 | LOV_MAGIC_MAGIC)
1183 #define LOV_MAGIC_SEL (0x0BD80000 | LOV_MAGIC_MAGIC)
1186 * magic for fully defined striping
1187 * the idea is that we should have different magics for striping "hints"
1188 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1189 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1190 * we can't just change it w/o long way preparation, but we still need a
1191 * mechanism to allow LOD to differentiate hint versus ready striping.
1192 * so, at the moment we do a trick: MDT knows what to expect from request
1193 * depending on the case (replay uses ready striping, non-replay req uses
1194 * hints), so MDT replaces magic with appropriate one and now LOD can
1195 * easily understand what's inside -bzzz
1197 * those *_DEF magics are only used on server side internally, they
1198 * won't be put on wire or disk.
1200 #define LOV_MAGIC_DEFINED 0x10000000
1201 #define LOV_MAGIC_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V1)
1202 #define LOV_MAGIC_V3_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V3)
1203 #define LOV_MAGIC_COMP_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_COMP_V1)
1205 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1206 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1208 #define lov_ost_data lov_ost_data_v1
1209 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1210 struct ost_id l_ost_oi; /* OST object ID */
1211 __u32 l_ost_gen; /* generation of this l_ost_idx */
1212 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1215 #define lov_mds_md lov_mds_md_v1
1216 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1217 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1218 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1219 struct ost_id lmm_oi; /* LOV object ID */
1220 __u32 lmm_stripe_size; /* size of stripe in bytes */
1221 /* lmm_stripe_count used to be __u32 */
1222 __u16 lmm_stripe_count; /* num stripes in use for this object */
1223 __u16 lmm_layout_gen; /* layout generation number */
1224 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1227 #define MAX_MD_SIZE_OLD (sizeof(struct lov_mds_md) + \
1228 4 * sizeof(struct lov_ost_data))
1229 #define MAX_MD_SIZE (sizeof(struct lov_comp_md_v1) + \
1230 4 * (sizeof(struct lov_comp_md_entry_v1) + \
1232 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * \
1233 sizeof(struct lov_ost_data))
1235 /* default MDT reply size allocated, should the striping be bigger, it will be
1236 * reallocated in mdt_fix_reply. 100 stipes ~= 2.5KB of data.
1238 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1239 100 * sizeof(struct lov_ost_data))
1241 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1242 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1243 #define XATTR_USER_PREFIX "user."
1244 #define XATTR_TRUSTED_PREFIX "trusted."
1245 #define XATTR_SECURITY_PREFIX "security."
1246 #define XATTR_ENCRYPTION_PREFIX "encryption."
1248 #define XATTR_NAME_SOM "trusted.som"
1249 #define XATTR_NAME_LOV "trusted.lov"
1250 #define XATTR_NAME_LMA "trusted.lma"
1251 #define XATTR_NAME_LMV "trusted.lmv"
1252 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1253 #define XATTR_NAME_LINK "trusted.link"
1254 #define XATTR_NAME_FID "trusted.fid"
1255 #define XATTR_NAME_VERSION "trusted.version"
1256 #define XATTR_NAME_SOM "trusted.som"
1257 #define XATTR_NAME_HSM "trusted.hsm"
1258 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1259 #define XATTR_NAME_DUMMY "trusted.dummy"
1260 #define XATTR_NAME_PROJID "trusted.projid"
1261 #define XATTR_NAME_DATAVER "trusted.dataver"
1263 #define LL_XATTR_NAME_ENCRYPTION_CONTEXT_OLD XATTR_SECURITY_PREFIX"c"
1264 #define LL_XATTR_NAME_ENCRYPTION_CONTEXT XATTR_ENCRYPTION_PREFIX"c"
1266 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1267 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1269 #define XATTR_NAME_JOB_DEFAULT "user.job"
1270 /* longest allowed jobid xattr name is "user." + 7 chars + null terminator */
1271 #define XATTR_JOB_MAX_LEN 13
1273 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1274 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1275 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1276 struct ost_id lmm_oi; /* LOV object ID */
1277 __u32 lmm_stripe_size; /* size of stripe in bytes */
1278 /* lmm_stripe_count used to be __u32 */
1279 __u16 lmm_stripe_count; /* num stripes in use for this object */
1280 __u16 lmm_layout_gen; /* layout generation number */
1281 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1282 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1285 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1287 if (stripes == LOV_ALL_STRIPES)
1290 if (lmm_magic == LOV_MAGIC_V1)
1291 return sizeof(struct lov_mds_md_v1) +
1292 stripes * sizeof(struct lov_ost_data_v1);
1293 else /* LOV_MAGIC_V3 and LOV_MAGIC_SPECIFIC */
1294 return sizeof(struct lov_mds_md_v3) +
1295 stripes * sizeof(struct lov_ost_data_v1);
1299 lov_mds_md_max_stripe_count(__kernel_size_t buf_size, __u32 lmm_magic)
1301 switch (lmm_magic) {
1302 case LOV_MAGIC_V1: {
1303 struct lov_mds_md_v1 lmm;
1305 if (buf_size < sizeof(lmm))
1308 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1310 case LOV_MAGIC_V3: {
1311 struct lov_mds_md_v3 lmm;
1313 if (buf_size < sizeof(lmm))
1316 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1323 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1324 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1325 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1326 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1327 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1328 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1329 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1330 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1331 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1332 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1333 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1334 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1335 #define OBD_MD_DOM_SIZE (0x00001000ULL) /* Data-on-MDT component size */
1336 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1337 #define OBD_MD_FLPARENT (0x00004000ULL) /* parent FID */
1338 #define OBD_MD_LAYOUT_VERSION (0x00008000ULL) /* OST object layout version */
1339 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1340 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1341 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1342 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1343 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1344 /* OBD_MD_FLQOS (0x00200000ULL) has never been used */
1345 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1346 #define OBD_MD_FLPRJQUOTA (0x00400000ULL) /* over quota flags sent from ost */
1347 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1348 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1349 /* OBD_MD_FLEPOCH (0x04000000ULL) obsolete 2.7.50 */
1350 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1351 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1352 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1353 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1354 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1356 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1357 /* OBD_MD_REINT (0x0000000200000000ULL) obsolete 1.8 */
1358 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1359 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1361 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1362 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1363 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1364 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1365 #define OBD_MD_FLAGSTATFS (0x0000010000000000ULL) /* aggregated statfs */
1366 /* OBD_MD_FLMDSCAPA (0x0000020000000000ULL) obsolete 2.7.54 */
1367 /* OBD_MD_FLOSSCAPA (0x0000040000000000ULL) obsolete 2.7.54 */
1368 /* OBD_MD_FLCKSPLIT (0x0000080000000000ULL) obsolete 2.3.58*/
1369 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1370 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1371 * under lock; for xattr
1372 * requests means the
1373 * client holds the lock
1375 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1377 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1378 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* intent executed */
1380 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1381 #define OBD_MD_FLOSTLAYOUT (0x0080000000000000ULL) /* contain ost_layout */
1382 #define OBD_MD_FLPROJID (0x0100000000000000ULL) /* project ID */
1383 #define OBD_MD_SECCTX (0x0200000000000000ULL) /* embed security xattr */
1384 #define OBD_MD_FLLAZYSIZE (0x0400000000000000ULL) /* Lazy size */
1385 #define OBD_MD_FLLAZYBLOCKS (0x0800000000000000ULL) /* Lazy blocks */
1386 #define OBD_MD_FLBTIME (0x1000000000000000ULL) /* birth time */
1387 #define OBD_MD_ENCCTX (0x2000000000000000ULL) /* embed encryption ctx */
1388 #define OBD_MD_NAMEHASH (0x4000000000000000ULL) /* use hash instead of name
1389 * in case of encryption
1392 #define OBD_MD_FLALLQUOTA (OBD_MD_FLUSRQUOTA | \
1393 OBD_MD_FLGRPQUOTA | \
1396 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1397 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1398 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1399 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1400 OBD_MD_FLPARENT | OBD_MD_FLRDEV | OBD_MD_FLGROUP | \
1401 OBD_MD_FLPROJID | OBD_MD_FLBTIME)
1403 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1405 /* don't forget obdo_fid which is way down at the bottom so it can
1406 * come after the definition of llog_cookie
1411 HSS_CLEARMASK = 0x02,
1412 HSS_ARCHIVE_ID = 0x04,
1415 struct hsm_state_set {
1417 __u32 hss_archive_id;
1419 __u64 hss_clearmask;
1422 /* ost_body.data values for OST_BRW */
1424 #define OBD_BRW_READ 0x01
1425 #define OBD_BRW_WRITE 0x02
1426 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1427 #define OBD_BRW_NDELAY 0x04 /* Non-delay RPC should be issued for
1428 * this page. Non-delay RPCs have bit
1431 #define OBD_BRW_SYNC 0x08 /* part of sync transfer, not in grant */
1432 #define OBD_BRW_CHECK 0x10
1433 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1434 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1435 /* OBD_BRW_NOCACHE is currently neither set nor tested */
1436 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1437 #define OBD_BRW_NOQUOTA 0x100 /* do not enforce quota */
1438 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1439 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1440 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1441 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1442 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1443 #define OBD_BRW_SOFT_SYNC 0x4000 /* flag notifies server that client is
1444 * running low on space for unstable
1445 * pages; asking it to sync quickly
1447 #define OBD_BRW_OVER_PRJQUOTA 0x8000 /* Running out of project quota */
1448 #define OBD_BRW_ROOT_PRJQUOTA 0x10000 /* check project quota for root */
1449 #define OBD_BRW_RDMA_ONLY 0x20000 /* RPC contains RDMA-only pages*/
1450 #define OBD_BRW_SYS_RESOURCE 0x40000 /* page has CAP_SYS_RESOURCE */
1451 #define OBD_BRW_COMPRESSED 0x80000 /* data compressed on client */
1453 #define OBD_BRW_OVER_ALLQUOTA (OBD_BRW_OVER_USRQUOTA | \
1454 OBD_BRW_OVER_GRPQUOTA | \
1455 OBD_BRW_OVER_PRJQUOTA)
1457 #define OBD_BRW_DONE 0x40000000UL /*
1458 * osd-ldiskfs inernal,
1459 * IO has been issued before
1461 #define OBD_BRW_LOCAL1 0x80000000UL /*
1462 * osd-ldiskfs internal,
1463 * page mapped to real block
1466 #define OBD_BRW_LOCALS (OBD_BRW_LOCAL1 | OBD_BRW_DONE)
1468 #define OBD_MAX_GRANT 0x7fffffffUL /* Max grant allowed to one client: 2 GiB */
1470 #define OBD_OBJECT_EOF LUSTRE_EOF
1472 #define OST_MIN_PRECREATE 32
1473 #define OST_MAX_PRECREATE 20000
1476 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1477 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1478 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1479 * high 16 bits in 2.4 and later
1481 __u32 ioo_bufcnt; /* number of niobufs for this object */
1484 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1485 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1486 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits.
1488 #define IOOBJ_MAX_BRW_BITS 16
1489 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1490 #define ioobj_max_brw_set(ioo, num) \
1491 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1493 /* multiple of 8 bytes => can array */
1494 struct niobuf_remote {
1500 /* lock value block communicated between the filter and llite */
1502 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1503 * negative, i.e. because ((MASK + rc) & MASK) != MASK.
1505 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1506 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1507 #define OST_LVB_IS_ERR(blocks) \
1508 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1509 #define OST_LVB_SET_ERR(blocks, rc) blocks = OST_LVB_ERR_INIT + rc
1510 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1533 * lquota data structures
1536 /* The lquota_id structure is an union of all the possible identifier types that
1537 * can be used with quota, this includes:
1540 * - a FID which can be used for per-directory quota in the future
1543 struct lu_fid qid_fid; /* FID for per-directory quota */
1544 __u64 qid_uid; /* user identifier */
1545 __u64 qid_gid; /* group identifier */
1546 __u64 qid_projid; /* project identifier */
1549 /* quotactl management */
1550 struct obd_quotactl {
1552 __u32 qc_type; /* see Q_* flag below */
1555 struct obd_dqinfo qc_dqinfo;
1556 struct obd_dqblk qc_dqblk;
1560 #define qc_iter_md_offset qc_dqblk.dqb_bhardlimit
1561 #define qc_iter_dt_offset qc_dqblk.dqb_ihardlimit
1562 #define qc_iter_md_buflen qc_dqblk.dqb_bsoftlimit
1563 #define qc_iter_dt_buflen qc_dqblk.dqb_isoftlimit
1564 #define qc_iter_list qc_dqblk.dqb_curspace
1565 #define qc_iter_qid_start qc_dqblk.dqb_curinodes
1566 #define qc_iter_qid_end qc_dqblk.dqb_btime
1568 #define Q_COPY(out, in, member) (out)->member = (in)->member
1570 #define QCTL_COPY_NO_PNAME(out, in) \
1572 Q_COPY(out, in, qc_cmd); \
1573 Q_COPY(out, in, qc_type); \
1574 Q_COPY(out, in, qc_id); \
1575 Q_COPY(out, in, qc_stat); \
1576 Q_COPY(out, in, qc_dqinfo); \
1577 Q_COPY(out, in, qc_dqblk); \
1581 * - in and out maybe a type of struct if_quotactl or struct obd_quotactl
1582 * - in and out need not be of the same type.
1584 #define QCTL_COPY(out, in) \
1586 QCTL_COPY_NO_PNAME(out, in); \
1587 if (LUSTRE_Q_CMD_IS_POOL(in->qc_cmd)) { \
1588 size_t len = strnlen(in->qc_poolname, LOV_MAXPOOLNAME); \
1590 memcpy(out->qc_poolname, in->qc_poolname, len); \
1591 out->qc_poolname[len] = '\0'; \
1595 /* Body of quota request used for quota acquire/release RPCs between quota
1596 * master (aka QMT) and slaves (ak QSD).
1599 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1600 * and type (data or metadata) as well as
1601 * the quota type (user or group).
1603 union lquota_id qb_id; /* uid or gid or directory FID */
1604 __u32 qb_flags; /* see below */
1606 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1607 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1608 __u64 qb_slv_ver; /* slave index file version */
1609 struct lustre_handle qb_lockh; /* per-ID lock handle */
1610 struct lustre_handle qb_glb_lockh; /* global lock handle */
1611 __u64 qb_padding1[4];
1614 /* When the quota_body is used in the reply of quota global intent
1615 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID.
1617 #define qb_slv_fid qb_fid
1618 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1621 #define qb_qunit qb_usage
1623 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1624 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1625 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1626 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1628 /* Quota types currently supported */
1630 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1631 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1632 LQUOTA_TYPE_PRJ = 0x02, /* maps to PRJQUOTA */
1636 /* There are 2 different resource types on which a quota limit can be enforced:
1637 * - inodes on the MDTs
1638 * - blocks on the OSTs
1641 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1642 LQUOTA_RES_DT = 0x02,
1644 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1646 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1649 * Space accounting support
1650 * Format of an accounting record, providing disk usage information for a given
1653 struct lquota_acct_rec { /* 16 bytes */
1654 __u64 bspace; /* current space in use */
1655 __u64 ispace; /* current # inodes in use */
1659 * Global quota index support
1660 * Format of a global record, providing global quota settings for a given quota
1663 struct lquota_glb_rec { /* 32 bytes */
1664 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1665 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1666 __u64 qbr_time; /* grace time, in seconds */
1667 __u64 qbr_granted; /* amount granted to slaves (inodes or KiB) */
1671 * Slave index support
1672 * Format of a slave record, recording how much space is granted to a given
1675 struct lquota_slv_rec { /* 8 bytes */
1676 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1677 * in #inodes or kbytes
1681 /* Data structures associated with the quota locks */
1683 /* Glimpse descriptor used for the index & per-ID quota locks */
1684 struct ldlm_gl_lquota_desc {
1685 union lquota_id gl_id; /* quota ID subject to the glimpse */
1686 __u64 gl_flags; /* see LQUOTA_FL* below */
1687 __u64 gl_ver; /* new index version */
1688 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1689 __u64 gl_softlimit; /* new softlimit */
1693 #define gl_qunit gl_hardlimit /* current qunit value used when
1694 * glimpsing per-ID quota locks
1697 /* quota glimpse flags */
1698 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1700 /* LVB used with quota (global and per-ID) locks */
1702 __u64 lvb_flags; /* see LQUOTA_FL* above */
1703 __u64 lvb_id_may_rel; /* space that might be released later */
1704 __u64 lvb_id_rel; /* space released by the slave for this ID */
1705 __u64 lvb_id_qunit; /* current qunit value */
1709 /* LVB used with global quota lock */
1710 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1718 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1727 MDS_GETATTR_NAME = 34,
1732 MDS_DISCONNECT = 39,
1735 MDS_PIN = 42, /* obsolete, never used in a release */
1736 MDS_UNPIN = 43, /* obsolete, never used in a release */
1738 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1740 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1743 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1745 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1747 MDS_HSM_STATE_GET = 54,
1748 MDS_HSM_STATE_SET = 55,
1749 MDS_HSM_ACTION = 56,
1750 MDS_HSM_PROGRESS = 57,
1751 MDS_HSM_REQUEST = 58,
1752 MDS_HSM_CT_REGISTER = 59,
1753 MDS_HSM_CT_UNREGISTER = 60,
1754 MDS_SWAP_LAYOUTS = 61,
1760 #define MDS_FIRST_OPC MDS_GETATTR
1763 /* opcodes for object update */
1769 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1789 /* the disposition of the intent outlines what was executed */
1790 #define DISP_IT_EXECD 0x00000001
1791 #define DISP_LOOKUP_EXECD 0x00000002
1792 #define DISP_LOOKUP_NEG 0x00000004
1793 #define DISP_LOOKUP_POS 0x00000008
1794 #define DISP_OPEN_CREATE 0x00000010
1795 #define DISP_OPEN_OPEN 0x00000020
1796 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1797 #define DISP_ENQ_OPEN_REF 0x00800000
1798 #define DISP_ENQ_CREATE_REF 0x01000000
1799 #define DISP_OPEN_LOCK 0x02000000
1800 #define DISP_OPEN_LEASE 0x04000000
1801 #define DISP_OPEN_STRIPE 0x08000000
1802 #define DISP_OPEN_DENY 0x10000000
1804 /* INODE LOCK PARTS */
1805 enum mds_ibits_locks {
1806 MDS_INODELOCK_LOOKUP = 0x000001, /* For namespace, dentry etc. Was
1807 * used to protect permission (mode,
1808 * owner, group, etc) before 2.4.
1810 MDS_INODELOCK_UPDATE = 0x000002, /* size, links, timestamps */
1811 MDS_INODELOCK_OPEN = 0x000004, /* For opened files */
1812 MDS_INODELOCK_LAYOUT = 0x000008, /* for layout */
1814 /* The PERM bit is added in 2.4, and is used to protect permission
1815 * (mode, owner, group, ACL, etc.) separate from LOOKUP lock.
1816 * For remote directories (in DNE) these locks will be granted by
1817 * different MDTs (different LDLM namespace).
1819 * For local directory, the MDT always grants UPDATE|PERM together.
1820 * For remote directory, master MDT (where remote directory is) grants
1821 * UPDATE|PERM, and remote MDT (where name entry is) grants LOOKUP_LOCK.
1823 MDS_INODELOCK_PERM = 0x000010,
1824 MDS_INODELOCK_XATTR = 0x000020, /* non-permission extended attrs */
1825 MDS_INODELOCK_DOM = 0x000040, /* Data for Data-on-MDT files */
1826 /* Do not forget to increase MDS_INODELOCK_NUMBITS when adding bits */
1828 #define MDS_INODELOCK_NUMBITS 7
1829 /* This FULL lock is useful to take on unlink sort of operations */
1830 #define MDS_INODELOCK_FULL ((1 << MDS_INODELOCK_NUMBITS) - 1)
1831 /* DOM lock shouldn't be canceled early, use this macro for ELC */
1832 #define MDS_INODELOCK_ELC (MDS_INODELOCK_FULL & ~MDS_INODELOCK_DOM)
1834 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1835 * but was moved into name[1] along with the OID to avoid consuming the
1836 * name[2,3] fields that need to be used for the quota id (also a FID).
1839 LUSTRE_RES_ID_SEQ_OFF = 0,
1840 LUSTRE_RES_ID_VER_OID_OFF = 1,
1841 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1842 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1843 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1844 LUSTRE_RES_ID_HSH_OFF = 3
1847 #define MDS_STATUS_CONN 1
1848 #define MDS_STATUS_LOV 2
1851 /* these should be identical to their EXT4_*_FL counterparts, they are
1852 * redefined here only to avoid dragging in fs/ext4/ext4.h
1854 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1855 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1856 LUSTRE_APPEND_FL = 0x00000020, /* file writes may only append */
1857 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1858 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1859 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1860 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1861 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1862 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1863 LUSTRE_PROJINHERIT_FL = 0x20000000, /* Create with parents projid */
1865 /* These flags will not be identical to any EXT4_*_FL counterparts,
1866 * and only reserved for lustre purpose. Note: these flags might
1867 * be conflict with some of EXT4 flags, so
1868 * 1. these conflict flags needs to be removed when the flag is
1869 * wired by la_flags see osd_attr_get().
1870 * 2. If these flags needs to be stored into inode, they will be
1871 * stored in LMA. see LMAI_XXXX
1873 LUSTRE_ORPHAN_FL = 0x00002000,
1874 LUSTRE_SET_SYNC_FL = 0x00040000, /* Synchronous setattr on OSTs */
1875 LUSTRE_ENCRYPT_FL = 0x00800000, /* encrypted file */
1877 LUSTRE_LMA_FL_MASKS = LUSTRE_ENCRYPT_FL | LUSTRE_ORPHAN_FL,
1880 #ifndef FS_XFLAG_SYNC
1881 #define FS_XFLAG_SYNC 0x00000020 /* all writes synchronous */
1883 #ifndef FS_XFLAG_NOATIME
1884 #define FS_XFLAG_NOATIME 0x00000040 /* do not update access time */
1886 #ifndef FS_XFLAG_IMMUTABLE
1887 #define FS_XFLAG_IMMUTABLE 0x00000008 /* file cannot be modified */
1889 #ifndef FS_XFLAG_APPEND
1890 #define FS_XFLAG_APPEND 0x00000010 /* all writes append */
1892 #ifndef FS_XFLAG_PROJINHERIT
1893 #define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
1896 /* 64 possible states */
1897 enum md_transient_state {
1898 MS_RESTORE = (1 << 0), /* restore is running */
1902 struct lu_fid mbo_fid1;
1903 struct lu_fid mbo_fid2;
1904 struct lustre_handle mbo_open_handle;
1906 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1910 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1911 __u64 mbo_version; /* was mbo_ioepoch before 2.11 */
1912 __u64 mbo_t_state; /* transient file state defined in
1913 * enum md_transient_state
1914 * was "ino" until 2.4.0
1918 __u32 mbo_capability;
1922 __u32 mbo_flags; /* most replies: LUSTRE_*_FL file attributes,
1923 * data_version: OBD_FL_* flags
1926 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1927 __u32 mbo_layout_gen; /* was "generation" until 2.4.0 */
1929 __u32 mbo_eadatasize;
1931 __u32 mbo_max_mdsize;
1932 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1933 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1934 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1936 __u64 mbo_dom_size; /* size of DOM component */
1937 __u64 mbo_dom_blocks; /* blocks consumed by DOM component */
1939 __u64 mbo_padding_9; /* also fix lustre_swab_mdt_body */
1940 __u64 mbo_padding_10;
1943 struct mdt_ioepoch {
1944 struct lustre_handle mio_open_handle;
1945 __u64 mio_unused1; /* was ioepoch */
1946 __u32 mio_unused2; /* was flags */
1950 /* permissions for md_perm.mp_perm */
1952 CFS_SETUID_PERM = 0x01,
1953 CFS_SETGID_PERM = 0x02,
1954 CFS_SETGRP_PERM = 0x04,
1957 struct mdt_rec_setattr {
1967 __u32 sa_padding_1_h;
1968 struct lu_fid sa_fid;
1977 __u32 sa_attr_flags;
1979 __u32 sa_bias; /* some operation flags */
1986 * Attribute flags used in mdt_rec_setattr::sa_valid.
1987 * The kernel's #defines for ATTR_* should not be used over the network
1988 * since the client and MDS may run different kernels (see bug 13828)
1989 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1991 enum mds_attr_flags {
1992 MDS_ATTR_MODE = 0x1ULL, /* = 1 */
1993 MDS_ATTR_UID = 0x2ULL, /* = 2 */
1994 MDS_ATTR_GID = 0x4ULL, /* = 4 */
1995 MDS_ATTR_SIZE = 0x8ULL, /* = 8 */
1996 MDS_ATTR_ATIME = 0x10ULL, /* = 16 */
1997 MDS_ATTR_MTIME = 0x20ULL, /* = 32 */
1998 MDS_ATTR_CTIME = 0x40ULL, /* = 64 */
1999 MDS_ATTR_ATIME_SET = 0x80ULL, /* = 128 */
2000 MDS_ATTR_MTIME_SET = 0x100ULL, /* = 256 */
2001 MDS_ATTR_FORCE = 0x200ULL, /* = 512, change it */
2002 MDS_ATTR_ATTR_FLAG = 0x400ULL, /* = 1024 */
2003 MDS_ATTR_KILL_SUID = 0x800ULL, /* = 2048 */
2004 MDS_ATTR_KILL_SGID = 0x1000ULL, /* = 4096 */
2005 MDS_ATTR_CTIME_SET = 0x2000ULL, /* = 8192 */
2006 MDS_ATTR_FROM_OPEN = 0x4000ULL, /* = 16384, from open O_TRUNC */
2007 MDS_ATTR_BLOCKS = 0x8000ULL, /* = 32768 */
2008 MDS_ATTR_PROJID = 0x10000ULL, /* = 65536 */
2009 MDS_ATTR_LSIZE = 0x20000ULL, /* = 131072 */
2010 MDS_ATTR_LBLOCKS = 0x40000ULL, /* = 262144 */
2011 MDS_ATTR_OVERRIDE = 0x2000000ULL, /* = 33554432 */
2015 /* MDS_CHECK_SPLIT = 1 << 0, obsolete before 2.3.58 */
2016 /* used for remote object getattr/open by name: in the original
2017 * getattr/open request, MDT found the object against name is on another
2018 * MDT, then packed FID and LOOKUP lock in reply and returned -EREMOTE,
2019 * and client knew it's a remote object, then set this flag in
2020 * getattr/open request and sent to the corresponding MDT to finish
2021 * getattr/open, which fetched attributes and UPDATE lock/opened file.
2023 MDS_CROSS_REF = 1 << 1,
2024 /* MDS_VTX_BYPASS = 1 << 2, obsolete since 2.3.54 */
2025 MDS_PERM_BYPASS = 1 << 3,
2026 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
2027 MDS_QUOTA_IGNORE = 1 << 5,
2028 /* MDS_CLOSE_CLEANUP = 1 << 6, obsolete since 2.3.51 */
2029 MDS_KEEP_ORPHAN = 1 << 7,
2030 MDS_RECOV_OPEN = 1 << 8,
2031 MDS_DATA_MODIFIED = 1 << 9,
2032 MDS_CREATE_VOLATILE = 1 << 10,
2033 MDS_OWNEROVERRIDE = 1 << 11,
2034 MDS_HSM_RELEASE = 1 << 12,
2035 MDS_CLOSE_MIGRATE = 1 << 13,
2036 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
2037 MDS_CLOSE_LAYOUT_MERGE = 1 << 15,
2038 MDS_CLOSE_RESYNC_DONE = 1 << 16,
2039 MDS_CLOSE_LAYOUT_SPLIT = 1 << 17,
2040 MDS_TRUNC_KEEP_LEASE = 1 << 18,
2041 MDS_PCC_ATTACH = 1 << 19,
2042 MDS_CLOSE_UPDATE_TIMES = 1 << 20,
2043 /* setstripe create only, don't restripe if target exists */
2044 MDS_SETSTRIPE_CREATE = 1 << 21,
2045 MDS_FID_OP = 1 << 22,
2046 /* migrate dirent only */
2047 MDS_MIGRATE_NSONLY = 1 << 23,
2048 /* create with default LMV from client */
2049 MDS_CREATE_DEFAULT_LMV = 1 << 24,
2050 /* Compat flag with clients that do not send old and new data version
2051 * after swap layout */
2052 MDS_CLOSE_LAYOUT_SWAP_HSM = 1 << 25,
2055 #define MDS_CLOSE_INTENT (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP | \
2056 MDS_CLOSE_LAYOUT_MERGE | MDS_CLOSE_LAYOUT_SPLIT | \
2057 MDS_CLOSE_RESYNC_DONE)
2059 /* instance of mdt_reint_rec */
2060 struct mdt_rec_create {
2068 __u32 cr_suppgid1_h;
2070 __u32 cr_suppgid2_h;
2071 struct lu_fid cr_fid1;
2072 struct lu_fid cr_fid2;
2073 struct lustre_handle cr_open_handle_old; /* in case of open replay */
2077 __u32 cr_archive_id;
2080 __u64 cr_padding_1; /* rr_blocks */
2083 /* use of helpers set/get_mrc_cr_flags() is needed to access
2084 * 64 bits cr_flags [cr_flags_l, cr_flags_h]
2086 __u32 cr_flags_l; /* for use with open, low 32 bits */
2087 __u32 cr_flags_h; /* for use with open, high 32 bits */
2088 __u32 cr_umask; /* umask for create */
2089 __u32 cr_padding_4; /* rr_padding_4 */
2092 /* instance of mdt_reint_rec */
2093 struct mdt_rec_link {
2101 __u32 lk_suppgid1_h;
2103 __u32 lk_suppgid2_h;
2104 struct lu_fid lk_fid1;
2105 struct lu_fid lk_fid2;
2107 __u64 lk_padding_1; /* rr_atime */
2108 __u64 lk_padding_2; /* rr_ctime */
2109 __u64 lk_padding_3; /* rr_size */
2110 __u64 lk_padding_4; /* rr_blocks */
2112 __u32 lk_padding_5; /* rr_mode */
2113 __u32 lk_padding_6; /* rr_flags */
2114 __u32 lk_padding_7; /* rr_padding_2 */
2115 __u32 lk_padding_8; /* rr_padding_3 */
2116 __u32 lk_padding_9; /* rr_padding_4 */
2119 /* instance of mdt_reint_rec */
2120 struct mdt_rec_unlink {
2128 __u32 ul_suppgid1_h;
2130 __u32 ul_suppgid2_h;
2131 struct lu_fid ul_fid1;
2132 struct lu_fid ul_fid2;
2134 __u64 ul_padding_2; /* rr_atime */
2135 __u64 ul_padding_3; /* rr_ctime */
2136 __u64 ul_padding_4; /* rr_size */
2137 __u64 ul_padding_5; /* rr_blocks */
2140 __u32 ul_padding_6; /* rr_flags */
2141 __u32 ul_padding_7; /* rr_padding_2 */
2142 __u32 ul_padding_8; /* rr_padding_3 */
2143 __u32 ul_padding_9; /* rr_padding_4 */
2146 /* instance of mdt_reint_rec */
2147 struct mdt_rec_rename {
2155 __u32 rn_suppgid1_h;
2157 __u32 rn_suppgid2_h;
2158 struct lu_fid rn_fid1;
2159 struct lu_fid rn_fid2;
2161 __u64 rn_padding_1; /* rr_atime */
2162 __u64 rn_padding_2; /* rr_ctime */
2163 __u64 rn_padding_3; /* rr_size */
2164 __u64 rn_padding_4; /* rr_blocks */
2165 __u32 rn_bias; /* some operation flags */
2166 __u32 rn_mode; /* cross-ref rename has mode */
2167 __u32 rn_padding_5; /* rr_flags */
2168 __u32 rn_padding_6; /* rr_padding_2 */
2169 __u32 rn_padding_7; /* rr_padding_3 */
2170 __u32 rn_padding_8; /* rr_padding_4 */
2173 /* instance of mdt_reint_rec */
2174 struct mdt_rec_setxattr {
2182 __u32 sx_suppgid1_h;
2184 __u32 sx_suppgid2_h;
2185 struct lu_fid sx_fid;
2186 __u64 sx_padding_1; /* These three are rr_fid2 */
2191 __u64 sx_padding_5; /* rr_ctime */
2192 __u64 sx_padding_6; /* rr_size */
2193 __u64 sx_padding_7; /* rr_blocks */
2196 __u32 sx_padding_8; /* rr_flags */
2197 __u32 sx_padding_9; /* rr_padding_2 */
2198 __u32 sx_padding_10; /* rr_padding_3 */
2199 __u32 sx_padding_11; /* rr_padding_4 */
2202 /* instance of mdt_reint_rec. FLR: for file resync MDS_REINT_RESYNC RPC. */
2203 struct mdt_rec_resync {
2211 __u32 rs_suppgid1_h;
2213 __u32 rs_suppgid2_h;
2214 struct lu_fid rs_fid;
2215 __u8 rs_padding0[sizeof(struct lu_fid)];
2216 struct lustre_handle rs_lease_handle; /* rr_mtime */
2217 __s64 rs_padding1; /* rr_atime */
2218 __s64 rs_padding2; /* rr_ctime */
2219 __u64 rs_padding3; /* rr_size */
2220 __u64 rs_padding4; /* rr_blocks */
2222 __u32 rs_padding5; /* rr_mode */
2223 __u32 rs_padding6; /* rr_flags */
2224 __u32 rs_padding7; /* rr_flags_h */
2225 __u32 rs_padding8; /* rr_umask */
2227 __u16 rs_padding9; /* rr_padding_4 */
2231 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2232 * Do NOT change the size of various members, otherwise the value
2233 * will be broken in lustre_swab_mdt_rec_reint().
2235 * If you add new members in other mdt_reint_xxx structres and need to use the
2236 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2238 struct mdt_rec_reint {
2246 __u32 rr_suppgid1_h;
2248 __u32 rr_suppgid2_h;
2249 struct lu_fid rr_fid1;
2250 struct lu_fid rr_fid2;
2262 __u16 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2265 #define LMV_DESC_QOS_MAXAGE_DEFAULT 60 /* Seconds */
2267 /* lmv structures */
2269 __u32 ld_tgt_count; /* how many MDS's */
2270 __u32 ld_active_tgt_count; /* how many active */
2271 __u32 ld_default_stripe_count; /* how many objects are used */
2272 __u32 ld_pattern; /* default hash pattern */
2273 __u64 ld_default_hash_size;
2274 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2275 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2276 __u32 ld_qos_maxage; /* in second */
2277 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2278 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2279 struct obd_uuid ld_uuid;
2282 /* LMV layout EA, and it will be stored both in master and slave object */
2283 struct lmv_mds_md_v1 {
2285 __u32 lmv_stripe_count;
2286 __u32 lmv_master_mdt_index; /* On master object, it is master
2287 * MDT index, on slave object, it
2288 * is stripe index of the slave obj
2290 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2291 * which hash function to be used,
2292 * Note: only lower 16 bits is being
2293 * used for now. Higher 16 bits will
2294 * be used to mark the object status,
2295 * for example migrating or dead.
2297 __u32 lmv_layout_version; /* increased each time layout changed,
2298 * by directory migration, restripe
2301 __u32 lmv_migrate_offset; /* once this is set, it means this
2302 * directory is been migrated, stripes
2303 * before this offset belong to target,
2304 * from this to source.
2306 __u32 lmv_migrate_hash; /* hash type of source stripes of
2307 * migrating directory
2311 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
2312 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2315 /* stripe count before directory split */
2316 #define lmv_split_offset lmv_migrate_offset
2317 /* stripe count after directory merge */
2318 #define lmv_merge_offset lmv_migrate_offset
2319 /* directory hash type after merge */
2320 #define lmv_merge_hash lmv_migrate_hash
2322 /* foreign LMV EA */
2323 struct lmv_foreign_md {
2324 __u32 lfm_magic; /* magic number = LMV_MAGIC_FOREIGN */
2325 __u32 lfm_length; /* length of lfm_value */
2326 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
2327 __u32 lfm_flags; /* flags, type specific */
2328 char lfm_value[]; /* free format value */
2331 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2332 #define LMV_MAGIC LMV_MAGIC_V1
2334 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2335 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2336 #define LMV_MAGIC_FOREIGN 0x0CD50CD0 /* magic for lmv foreign */
2339 * The FNV-1a hash algorithm is as follows:
2340 * hash = FNV_offset_basis
2341 * for each octet_of_data to be hashed
2342 * hash = hash XOR octet_of_data
2343 * hash = hash × FNV_prime
2345 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2347 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2348 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2350 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2351 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2352 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, __kernel_size_t size)
2354 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2355 const unsigned char *p = buf;
2358 for (i = 0; i < size; i++) {
2360 hash *= LUSTRE_FNV_1A_64_PRIME;
2366 /* CRUSH placement group count */
2367 #define LMV_CRUSH_PG_COUNT 4096
2371 struct lmv_mds_md_v1 lmv_md_v1;
2372 struct lmv_user_md lmv_user_md;
2373 struct lmv_foreign_md lmv_foreign_md;
2376 static inline __kernel_ssize_t lmv_mds_md_size(int stripe_count,
2377 unsigned int lmm_magic)
2379 __kernel_ssize_t len = -EINVAL;
2381 switch (lmm_magic) {
2382 case LMV_MAGIC_V1: {
2383 struct lmv_mds_md_v1 *lmm1;
2385 len = sizeof(*lmm1);
2386 len += stripe_count * sizeof(lmm1->lmv_stripe_fids[0]);
2394 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2396 switch (__le32_to_cpu(lmm->lmv_magic)) {
2398 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2399 case LMV_USER_MAGIC:
2400 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2406 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2408 switch (__le32_to_cpu(lmm->lmv_magic)) {
2410 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2411 case LMV_USER_MAGIC:
2412 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2422 FLD_FIRST_OPC = FLD_QUERY
2428 SEQ_FIRST_OPC = SEQ_QUERY
2432 SEQ_ALLOC_SUPER = 0,
2444 LFSCK_NOTIFY = 1101,
2447 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2451 * LOV data structures
2454 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2455 /* The size of the buffer the lov/mdc reserves for the
2456 * array of UUIDs returned by the MDS. With the current
2457 * protocol, this will limit the max number of OSTs per LOV
2460 #define LOV_DESC_MAGIC 0xB0CCDE5C
2461 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2462 #define LOV_DESC_STRIPE_SIZE_DEFAULT (4 << LNET_MTU_BITS)
2464 /* LOV settings descriptor (should only contain static info) */
2466 __u32 ld_tgt_count; /* how many OBD's */
2467 __u32 ld_active_tgt_count; /* how many active */
2468 __s32 ld_default_stripe_count; /* how many objects are used */
2469 __u32 ld_pattern; /* default PATTERN_RAID0 */
2470 __u64 ld_default_stripe_size; /* in bytes */
2471 __s64 ld_default_stripe_offset; /* starting OST index */
2472 __u32 ld_padding_0; /* unused */
2473 __u32 ld_qos_maxage; /* in second */
2474 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2475 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2476 struct obd_uuid ld_uuid;
2479 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2484 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2489 LDLM_BL_CALLBACK = 104,
2490 LDLM_CP_CALLBACK = 105,
2491 LDLM_GL_CALLBACK = 106,
2492 LDLM_SET_INFO = 107,
2495 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2497 #define RES_NAME_SIZE 4
2498 struct ldlm_res_id {
2499 __u64 name[RES_NAME_SIZE];
2502 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2503 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2504 (unsigned long long)(res)->lr_name.name[1], \
2505 (unsigned long long)(res)->lr_name.name[2], \
2506 (unsigned long long)(res)->lr_name.name[3]
2523 #define LCK_MODE_NUM 9
2533 #define LDLM_MIN_TYPE LDLM_PLAIN
2535 struct ldlm_extent {
2541 static inline bool ldlm_extent_equal(const struct ldlm_extent *ex1,
2542 const struct ldlm_extent *ex2)
2544 return ex1->start == ex2->start && ex1->end == ex2->end;
2547 struct ldlm_inodebits {
2550 __u64 try_bits; /* optional bits to try */
2551 __u64 cancel_bits; /* for lock convert */
2555 __u32 li_initiator_id; /* index of MDT that initiated this lock */
2558 struct ldlm_flock_wire {
2566 /* it's important that the fields of the ldlm_extent structure match
2567 * the first fields of the ldlm_flock structure because there is only
2568 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2569 * this ever changes we will need to swab the union differently based
2570 * on the resource type.
2573 union ldlm_wire_policy_data {
2574 struct ldlm_extent l_extent;
2575 struct ldlm_flock_wire l_flock;
2576 struct ldlm_inodebits l_inodebits;
2579 struct barrier_lvb {
2585 struct ldlm_gl_barrier_desc {
2591 union ldlm_gl_desc {
2592 struct ldlm_gl_lquota_desc lquota_desc;
2593 struct ldlm_gl_barrier_desc barrier_desc;
2596 enum ldlm_intent_flags {
2597 IT_OPEN = 0x00000001,
2598 IT_CREAT = 0x00000002,
2599 IT_OPEN_CREAT = IT_OPEN | IT_CREAT, /* To allow case label. */
2600 IT_READDIR = 0x00000004, /* Used by mdc, not put on the wire. */
2601 IT_GETATTR = 0x00000008,
2602 IT_LOOKUP = 0x00000010,
2603 /* IT_UNLINK = 0x00000020, Obsolete. */
2604 /* IT_TRUNC = 0x00000040, Obsolete. */
2605 IT_GETXATTR = 0x00000080,
2606 /* IT_EXEC = 0x00000100, Obsolete. */
2607 /* IT_PIN = 0x00000200, Obsolete. */
2608 IT_LAYOUT = 0x00000400,
2609 IT_QUOTA_DQACQ = 0x00000800,
2610 IT_QUOTA_CONN = 0x00001000,
2611 /* IT_SETXATTR = 0x00002000, Obsolete. */
2612 IT_GLIMPSE = 0x00004000,
2613 IT_BRW = 0x00008000,
2616 struct ldlm_intent {
2620 struct ldlm_resource_desc {
2621 enum ldlm_type lr_type;
2622 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2623 struct ldlm_res_id lr_name;
2626 struct ldlm_lock_desc {
2627 struct ldlm_resource_desc l_resource;
2628 enum ldlm_mode l_req_mode;
2629 enum ldlm_mode l_granted_mode;
2630 union ldlm_wire_policy_data l_policy_data;
2633 #define LDLM_LOCKREQ_HANDLES 2
2634 #define LDLM_ENQUEUE_CANCEL_OFF 1
2636 struct ldlm_request {
2637 __u32 lock_flags; /* LDLM_FL_*,(lustre_dlm_flags.h) */
2638 __u32 lock_count; /* Num of locks in lock_handle[] */
2639 struct ldlm_lock_desc lock_desc; /* lock descriptor */
2640 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2645 __u32 lock_padding; /* fix lustre_swab_ldlm_reply */
2646 struct ldlm_lock_desc lock_desc;
2647 struct lustre_handle lock_handle;
2648 __u64 lock_policy_res1;
2649 __u64 lock_policy_res2;
2652 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2653 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2656 * Opcodes for mountconf (mgs and mgc)
2660 MGS_DISCONNECT = 251,
2661 MGS_EXCEPTION = 252, /* node died, etc. */
2662 MGS_TARGET_REG = 253, /* whenever target starts up */
2663 MGS_TARGET_DEL = 254,
2665 MGS_CONFIG_READ = 256,
2667 MGS_FIRST_OPC = MGS_CONNECT
2670 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 18, 53, 0)
2671 #define MGS_PARAM_MAXLEN 1024
2672 #define KEY_SET_INFO "set_info"
2674 struct mgs_send_param {
2675 char mgs_param[MGS_PARAM_MAXLEN];
2679 /* We pass this info to the MGS so it can write config logs */
2680 #define MTI_NAME_MAXLEN 64
2681 #define MTI_PARAM_MAXLEN 4096
2682 #define MTI_NIDS_MAX 32
2684 struct mgs_target_info {
2685 __u32 mti_lustre_ver;
2686 __u32 mti_stripe_index;
2687 __u32 mti_config_ver;
2688 __u32 mti_flags; /* LDD_F_* */
2689 __u32 mti_nid_count;
2690 __u32 mti_instance; /* Running instance of target */
2691 char mti_fsname[MTI_NAME_MAXLEN];
2692 char mti_svname[MTI_NAME_MAXLEN];
2693 char mti_uuid[sizeof(struct obd_uuid)];
2694 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t) */
2695 char mti_params[MTI_PARAM_MAXLEN];
2696 char mti_nidlist[][LNET_NIDSTR_SIZE];
2697 } __attribute__((packed));
2699 struct mgs_nidtbl_entry {
2700 __u64 mne_version; /* table version of this entry */
2701 __u32 mne_instance; /* target instance # */
2702 __u32 mne_index; /* target index */
2703 __u32 mne_length; /* length of this entry - by bytes */
2704 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2705 __u8 mne_nid_type; /* type of NID. for IPv6. */
2706 __u8 mne_nid_size; /* size of each NID, by bytes */
2707 __u8 mne_nid_count; /* # of NIDs in buffer */
2709 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2710 struct lnet_nid nidlist[0];
2715 MGS_CFG_T_CONFIG = 0,
2716 MGS_CFG_T_SPTLRPC = 1,
2717 MGS_CFG_T_RECOVER = 2,
2718 MGS_CFG_T_PARAMS = 3,
2719 MGS_CFG_T_NODEMAP = 4,
2720 MGS_CFG_T_BARRIER = 5,
2724 struct mgs_config_body {
2725 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2726 __u64 mcb_offset; /* next index of config log to request */
2727 __u16 mcb_type; /* type of log: MGS_CFG_T_[CONFIG|RECOVER] */
2729 __u8 mcb_nm_cur_pass; /* mcb_type == MGS_CFG_T_NODEMAP */
2730 __u8 mcb_rec_nid_size; /* mcb_type == MGS_CFG_T_RECOVER */
2732 __u8 mcb_bits; /* bits unit size of config log */
2733 __u32 mcb_units; /* # of units for bulk transfer */
2736 struct mgs_config_res {
2737 __u64 mcr_offset; /* index of last config log */
2739 __u64 mcr_size; /* size of the log */
2740 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2744 /* Config marker flags (in config log) */
2745 #define CM_START 0x01
2747 #define CM_SKIP 0x04
2748 #define CM_UPGRADE146 0x08
2749 #define CM_EXCLUDE 0x10
2750 #define CM_START_SKIP (CM_START | CM_SKIP)
2753 __u32 cm_step; /* aka config version */
2755 __u32 cm_vers; /* lustre release version number */
2756 __u32 cm_padding; /* 64 bit align */
2757 __s64 cm_createtime; /*when this record was first created */
2758 __s64 cm_canceltime; /*when this record is no longer valid*/
2759 char cm_tgtname[MTI_NAME_MAXLEN];
2760 char cm_comment[MTI_NAME_MAXLEN];
2764 * Opcodes for multiple servers.
2768 /* OBD_LOG_CANCEL = 401, obsolete since 1.5 */
2769 /* OBD_QC_CALLBACK = 402, obsolete since 2.4 */
2772 OBD_FIRST_OPC = OBD_PING
2776 * llog contexts indices.
2778 * There is compatibility problem with indexes below, they are not
2779 * continuous and must keep their numbers for compatibility needs.
2780 * See LU-5218 for details.
2783 LLOG_CONFIG_ORIG_CTXT = 0,
2784 LLOG_CONFIG_REPL_CTXT = 1,
2785 LLOG_MDS_OST_ORIG_CTXT = 2,
2786 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2787 LLOG_SIZE_ORIG_CTXT = 4,
2788 LLOG_SIZE_REPL_CTXT = 5,
2789 LLOG_TEST_ORIG_CTXT = 8,
2790 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2791 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2792 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2793 /* for multiple changelog consumers */
2794 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2795 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2796 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log. reserve for the client */
2797 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log. reserve for the client */
2801 /** Identifier for a single log object */
2803 struct ost_id lgl_oi;
2805 } __attribute__((packed));
2807 /** Records written to the CATALOGS list */
2808 #define CATLIST "CATALOGS"
2810 struct llog_logid lci_logid;
2814 } __attribute__((packed));
2816 /* Log data record types - there is no specific reason that these need to
2817 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2819 #define LLOG_OP_MAGIC 0x10600000
2820 #define LLOG_OP_MASK 0xfff00000
2823 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2824 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2825 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2826 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2827 REINT_UNLINK, /* obsolete after 2.5.0 */
2828 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2830 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2831 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2833 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2834 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2835 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2836 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2837 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2838 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2839 CHANGELOG_USER_REC2 = LLOG_OP_MAGIC | 0x70002,
2840 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2841 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000, /* Resevered to avoid
2844 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2845 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2848 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2849 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2851 /** Log record header - stored in little endian order.
2852 * Each record must start with this struct, end with a llog_rec_tail,
2853 * and be a multiple of 256 bits in size.
2855 struct llog_rec_hdr {
2860 } __attribute__((packed));
2862 struct llog_rec_tail {
2865 } __attribute__((packed));
2867 /* Where data follow just after header */
2868 #define REC_DATA(ptr) \
2869 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2871 #define REC_DATA_LEN(rec) \
2872 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2873 sizeof(struct llog_rec_tail))
2875 struct llog_logid_rec {
2876 struct llog_rec_hdr lid_hdr;
2877 struct llog_logid lid_id;
2881 struct llog_rec_tail lid_tail;
2882 } __attribute__((packed));
2884 struct llog_unlink_rec {
2885 struct llog_rec_hdr lur_hdr;
2889 struct llog_rec_tail lur_tail;
2890 } __attribute__((packed));
2892 struct llog_unlink64_rec {
2893 struct llog_rec_hdr lur_hdr;
2894 struct lu_fid lur_fid;
2895 __u32 lur_count; /* to destroy the lost precreated */
2899 struct llog_rec_tail lur_tail;
2900 } __attribute__((packed));
2902 struct llog_setattr64_rec {
2903 struct llog_rec_hdr lsr_hdr;
2904 struct ost_id lsr_oi;
2910 struct llog_rec_tail lsr_tail;
2911 } __attribute__((packed));
2913 /* Extended to support project quota */
2914 struct llog_setattr64_rec_v2 {
2915 struct llog_rec_hdr lsr_hdr;
2916 struct ost_id lsr_oi;
2923 __u32 lsr_layout_version;
2926 struct llog_rec_tail lsr_tail;
2927 } __attribute__((packed));
2929 struct llog_size_change_rec {
2930 struct llog_rec_hdr lsc_hdr;
2931 struct ll_fid lsc_fid;
2936 struct llog_rec_tail lsc_tail;
2937 } __attribute__((packed));
2939 #define CHANGELOG_MAGIC 0xca103000
2941 /** \a changelog_rec_type's that can't be masked */
2942 #define CHANGELOG_MINMASK BIT(CL_MARK)
2943 /** bits covering all \a changelog_rec_type's */
2944 #define CHANGELOG_ALLMASK (BIT(CL_LAST) - 1)
2945 /** default \a changelog_rec_type mask. Allow all of them, except
2946 * CL_ATIME since it can really be time consuming, and not necessary
2948 * Remove also CL_OPEN, CL_GETXATTR and CL_DN_OPEN from default list as it can
2949 * be costly and only necessary for audit purpose.
2951 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & \
2952 ~(BIT(CL_ATIME) | BIT(CL_OPEN) | BIT(CL_GETXATTR) | BIT(CL_DN_OPEN)))
2954 /* changelog llog name, needed by client replicators */
2955 #define CHANGELOG_CATALOG "changelog_catalog"
2957 struct changelog_setinfo {
2960 } __attribute__((packed));
2962 /** changelog record */
2963 struct llog_changelog_rec {
2964 struct llog_rec_hdr cr_hdr;
2965 struct changelog_rec cr; /**< Variable length field */
2966 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2967 } __attribute__((packed));
2969 #define CHANGELOG_USER_PREFIX "cl"
2970 #define CHANGELOG_USER_NAMELEN 16 /* base name including NUL terminator */
2971 #define CHANGELOG_USER_NAMELEN_FULL 30 /* basename plus 'cl$ID-' prefix */
2973 struct llog_changelog_user_rec {
2974 struct llog_rec_hdr cur_hdr;
2976 __u32 cur_time; /* relative time comparisons only */
2978 struct llog_rec_tail cur_tail;
2979 } __attribute__((packed));
2981 /* this is twice the size of CHANGELOG_USER_REC */
2982 struct llog_changelog_user_rec2 {
2983 struct llog_rec_hdr cur_hdr;
2985 __u32 cur_time; /* relative time comparisons only */
2989 char cur_name[CHANGELOG_USER_NAMELEN];
2992 struct llog_rec_tail cur_tail;
2993 } __attribute__((packed));
2995 enum agent_req_status {
3003 static inline const char *agent_req_status2name(enum agent_req_status ars)
3021 struct llog_agent_req_rec {
3022 struct llog_rec_hdr arr_hdr; /**< record header */
3023 __u32 arr_status; /**< status of the request */
3027 __u32 arr_archive_id; /**< backend archive number */
3028 __u64 arr_flags; /**< req flags */
3029 __u64 arr_compound_id; /** < compound cookie,
3032 __u64 arr_req_create; /**< req. creation time */
3033 __u64 arr_req_change; /**< req. status change time */
3034 struct hsm_action_item arr_hai; /**< req. to the agent */
3035 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
3036 } __attribute__((packed));
3038 /* Old llog gen for compatibility */
3042 } __attribute__((packed));
3044 struct llog_gen_rec {
3045 struct llog_rec_hdr lgr_hdr;
3046 struct llog_gen lgr_gen;
3050 struct llog_rec_tail lgr_tail;
3053 /* flags for the logs */
3055 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
3056 LLOG_F_IS_CAT = 0x2,
3057 LLOG_F_IS_PLAIN = 0x4,
3058 LLOG_F_EXT_JOBID = 0x8,
3059 LLOG_F_IS_FIXSIZE = 0x10,
3060 LLOG_F_EXT_EXTRA_FLAGS = 0x20,
3061 LLOG_F_EXT_X_UIDGID = 0x40,
3062 LLOG_F_EXT_X_NID = 0x80,
3063 LLOG_F_EXT_X_OMODE = 0x100,
3064 LLOG_F_EXT_X_XATTR = 0x200,
3065 LLOG_F_RM_ON_ERR = 0x400,
3066 LLOG_F_MAX_AGE = 0x800,
3068 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
3069 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
3070 * because the catlog record is usually fixed size, but its plain
3071 * log record can be variable
3073 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID | LLOG_F_EXT_EXTRA_FLAGS |
3074 LLOG_F_EXT_X_UIDGID | LLOG_F_EXT_X_NID |
3075 LLOG_F_EXT_X_OMODE | LLOG_F_EXT_X_XATTR,
3078 /* On-disk header structure of each log object, stored in little endian order */
3079 #define LLOG_MIN_CHUNK_SIZE 8192
3080 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
3081 * - sizeof(llh_bitmap)
3083 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
3084 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
3086 struct llog_log_hdr {
3087 struct llog_rec_hdr llh_hdr;
3088 __s64 llh_timestamp;
3090 __u32 llh_bitmap_offset;
3093 /* for a catalog the first/oldest and still in-use plain slot is just
3094 * next to it. It will serve as the upper limit after Catalog has
3098 struct obd_uuid llh_tgtuuid;
3099 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
3100 /* These fields must always be at the end of the llog_log_hdr.
3101 * Note: llh_bitmap size is variable because llog chunk size could be
3102 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
3103 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
3104 * llh_tail should only be refered by LLOG_HDR_TAIL().
3105 * But this structure is also used by client/server llog interface
3106 * (see llog_client.c), it will be kept in its original way to avoid
3107 * compatiblity issue.
3109 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
3110 struct llog_rec_tail llh_tail;
3111 } __attribute__((packed));
3112 #undef LLOG_HEADER_SIZE
3113 #undef LLOG_BITMAP_BYTES
3115 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
3116 llh->llh_bitmap_offset - \
3117 sizeof(llh->llh_tail)) * 8)
3118 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
3119 (llh)->llh_bitmap_offset)
3120 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
3121 llh->llh_hdr.lrh_len - \
3122 sizeof(llh->llh_tail)))
3124 /* log cookies are used to reference a specific log file and a record therein,
3125 * and pass record offset from llog_process_thread to llog_write
3127 struct llog_cookie {
3129 struct llog_logid lgc_lgl;
3135 } __attribute__((packed));
3137 /** llog protocol */
3138 enum llogd_rpc_ops {
3139 LLOG_ORIGIN_HANDLE_CREATE = 501,
3140 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
3141 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
3142 /* LLOG_ORIGIN_HANDLE_WRITE_REC = 504, Obsolete by 2.1. */
3143 /* LLOG_ORIGIN_HANDLE_CLOSE = 505, Obsolete by 1.8. */
3144 /* LLOG_ORIGIN_CONNECT = 506, Obsolete by 2.4. */
3145 /* LLOG_CATINFO = 507, Obsolete by 2.3. */
3146 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
3147 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* Obsolete by 2.11. */
3149 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
3153 struct llog_logid lgd_logid;
3155 __u32 lgd_llh_flags;
3157 __u32 lgd_saved_index;
3159 __u64 lgd_cur_offset;
3160 } __attribute__((packed));
3162 struct llogd_conn_body {
3163 struct llog_gen lgdc_gen;
3164 struct llog_logid lgdc_logid;
3165 __u32 lgdc_ctxt_idx;
3166 } __attribute__((packed));
3168 /* Note: 64-bit types are 64-bit aligned in structure */
3170 __u64 o_valid; /* hot fields in this obdo */
3173 __u64 o_size; /* o_size-o_blocks == ost_lvb */
3177 __u64 o_blocks; /* brw: cli sent cached bytes */
3180 /* 32-bit fields start here: keep an even number of them via padding */
3181 __u32 o_blksize; /* optimal IO blocksize */
3182 __u32 o_mode; /* brw: cli sent cache remain */
3186 __u32 o_nlink; /* brw: checksum */
3188 __u32 o_misc; /* brw: o_dropped */
3190 __u64 o_ioepoch; /* epoch in ost writes */
3191 __u32 o_stripe_idx; /* holds stripe idx */
3193 struct lustre_handle o_handle; /* brw: prolong locks on IO */
3194 /* Originally, the field is llog_cookie for destroy with unlink cookie
3195 * from MDS, it is obsolete in 2.8. Then reuse it by client to transfer
3196 * layout and PFL information in IO, setattr RPCs. Since llog_cookie is
3197 * not used on wire any longer, remove it from the obdo, then it can be
3198 * enlarged freely in the further without affect related RPCs.
3200 * sizeof(ost_layout) + sieof(__u32) == sizeof(llog_cookie).
3202 struct ost_layout o_layout;
3203 __u32 o_layout_version;
3207 __u64 o_data_version; /* getattr: sum of iversion for
3209 * brw: grant space consumed on
3210 * the client for the write
3213 __u32 o_padding_4; /* fix lustre_swab_obdo() */
3218 #define o_dirty o_blocks
3219 #define o_undirty o_mode
3220 #define o_dropped o_misc
3221 #define o_cksum o_nlink
3222 #define o_grant_used o_data_version
3223 #define o_falloc_mode o_nlink
3225 struct lfsck_request {
3237 __u16 lr_async_windows;
3239 struct lu_fid lr_fid;
3240 struct lu_fid lr_fid2;
3248 struct lfsck_reply {
3255 LE_LASTID_REBUILDING = 1,
3256 LE_LASTID_REBUILT = 2,
3262 /* LE_FID_ACCESSED = 8, moved to lfsck_events_local */
3264 LE_CONDITIONAL_DESTROY = 10,
3265 LE_PAIRS_VERIFY = 11,
3266 LE_SET_LMV_MASTER = 15,
3267 LE_SET_LMV_SLAVE = 16,
3270 enum lfsck_event_flags {
3271 LEF_TO_OST = 0x00000001,
3272 LEF_FROM_OST = 0x00000002,
3273 LEF_SET_LMV_HASH = 0x00000004,
3274 LEF_SET_LMV_ALL = 0x00000008,
3275 LEF_RECHECK_NAME_HASH = 0x00000010,
3276 LEF_QUERY_ALL = 0x00000020,
3279 /* request structure for OST's */
3284 /* Key for FIEMAP to be used in get_info calls */
3285 struct ll_fiemap_info_key {
3287 struct obdo lfik_oa;
3288 struct fiemap lfik_fiemap;
3291 #define IDX_INFO_MAGIC 0x3D37CC37
3293 /* List of flags used in idx_info::ii_flags */
3294 enum idx_info_flags {
3295 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3296 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3297 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3298 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3299 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
3302 /* Index file transfer through the network. The server serializes the index into
3303 * a byte stream which is sent to the client via a bulk transfer
3308 /* reply: see idx_info_flags below */
3309 __u32 ii_flags; /* II_FK_* flags */
3311 /* request & reply: number of lu_idxpage (to be) transferred */
3315 /* request: requested attributes passed down to the iterator API */
3318 /* request & reply: index file identifier (FID) */
3319 struct lu_fid ii_fid;
3321 /* reply: version of the index file before starting to walk the index.
3322 * Please note that the version can be modified at any time during the
3327 /* request: hash to start with:
3328 * reply: hash of the first entry of the first lu_idxpage and hash
3329 * of the entry to read next if any
3331 __u64 ii_hash_start;
3334 /* reply: keys size in lu_idxpages, minimal one if II_FL_VARKEY set */
3337 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC */
3345 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3347 #define LIP_MAGIC 0x8A6D6B6C
3349 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3351 /* 16-byte header */
3354 __u16 lip_nr; /* number of entries in the container */
3355 __u64 lip_pad0; /* additional padding for future use */
3357 /* key/record pairs are stored in the remaining 4080 bytes.
3358 * depending upon the flags in idx_info::ii_flags, each key/record
3359 * pair might be preceded by:
3361 * - the key size (II_FL_VARKEY is set)
3362 * - the record size (II_FL_VARREC is set)
3364 * For the time being, we only support fixed-size key & record.
3369 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3371 /* Gather all possible type associated with a 4KB container */
3373 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3374 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3375 char lp_array[LU_PAGE_SIZE];
3378 /* security opcodes */
3381 SEC_CTX_INIT_CONT = 802,
3384 SEC_FIRST_OPC = SEC_CTX_INIT,
3387 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3388 #define LINK_EA_MAGIC 0x11EAF1DFUL
3389 struct link_ea_header {
3392 __u64 leh_len; /* total size */
3393 __u32 leh_overflow_time;
3397 /** Hardlink data is name and parent fid.
3398 * Stored in this crazy struct for maximum packing and endian-neutrality
3400 struct link_ea_entry {
3401 unsigned char lee_reclen[2]; /* __u16 big-endian, unaligned */
3402 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3404 } __attribute__((packed));
3406 /** fid2path request/reply structure */
3407 struct getinfo_fid2path {
3408 struct lu_fid gf_fid;
3414 struct lu_fid gf_root_fid[0];
3416 } __attribute__((packed));
3418 /** path2parent request/reply structures */
3420 struct lu_fid gp_fid; /**< parent FID */
3421 __u32 gp_linkno; /**< hardlink number */
3422 __u32 gp_name_size; /**< size of the name field */
3423 char gp_name[0]; /**< zero-terminated link name */
3424 } __attribute__((packed));
3426 enum layout_intent_opc {
3427 LAYOUT_INTENT_ACCESS = 0, /** generic access */
3428 LAYOUT_INTENT_READ = 1, /** not used */
3429 LAYOUT_INTENT_WRITE = 2, /** write file, for comp layout */
3430 LAYOUT_INTENT_GLIMPSE = 3, /** not used */
3431 LAYOUT_INTENT_TRUNC = 4, /** truncate file, for comp layout */
3432 LAYOUT_INTENT_RELEASE = 5, /** reserved for HSM release */
3433 LAYOUT_INTENT_RESTORE = 6, /** reserved for HSM restore */
3434 LAYOUT_INTENT_PCCRO_SET = 7, /** set read-only layout for PCC */
3435 LAYOUT_INTENT_PCCRO_CLEAR = 8, /** clear read-only layout */
3438 /* enqueue layout lock with intent */
3439 struct layout_intent {
3440 __u32 lai_opc; /* intent operation for enqueue, read, write etc */
3442 struct lu_extent lai_extent;
3443 } __attribute__((packed));
3446 * On the wire version of hsm_progress structure.
3448 * Contains the userspace hsm_progress and some internal fields.
3450 struct hsm_progress_kernel {
3451 /* Field taken from struct hsm_progress */
3452 struct lu_fid hpk_fid;
3454 struct hsm_extent hpk_extent;
3456 __u16 hpk_errval; /* positive val */
3458 /* Additional fields */
3459 __u64 hpk_data_version;
3461 } __attribute__((packed));
3464 * OUT_UPDATE RPC Format
3466 * During the cross-ref operation, the Master MDT, which the client send the
3467 * request to, will disassembly the operation into object updates, then OSP
3468 * will send these updates to the remote MDT to be executed.
3470 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3471 * operation and does a type of modification to an object.
3479 * update (ub_count-th)
3481 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3486 * rc [+ buffers] (1st)
3487 * rc [+ buffers] (2st)
3489 * rc [+ buffers] (nr_count-th)
3491 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3492 * be equal to ub_count.
3496 * Type of each update, if adding/deleting update, please also update
3497 * update_opcode in lustre/target/out_lib.c.
3509 OUT_INDEX_LOOKUP = 9,
3510 OUT_INDEX_INSERT = 10,
3511 OUT_INDEX_DELETE = 11,
3517 OUT_XATTR_LIST = 17,
3522 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3523 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3524 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3525 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3528 struct object_update_param {
3529 __u16 oup_len; /* length of this parameter */
3533 } __attribute__((packed));
3536 struct object_update {
3537 __u16 ou_type; /* enum update_type */
3538 __u16 ou_params_count; /* update parameters count */
3539 __u32 ou_result_size; /* how many bytes can return */
3540 __u32 ou_flags; /* enum update_flag */
3541 __u32 ou_padding1; /* padding 1 */
3542 __u64 ou_batchid; /* op transno on master */
3543 struct lu_fid ou_fid; /* object to be updated */
3544 struct object_update_param ou_params[]; /* update params */
3547 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3548 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3549 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3550 /* Hold object_updates sending to the remote OUT in single RPC */
3551 struct object_update_request {
3553 __u16 ourq_count; /* number of ourq_updates[] */
3555 struct object_update ourq_updates[];
3558 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3559 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3560 /* Header for updates request between MDTs */
3561 struct out_update_header {
3564 __u32 ouh_inline_length;
3565 __u32 ouh_reply_size;
3566 __u32 ouh_inline_data[];
3569 struct out_update_buffer {
3574 /* the result of object update */
3575 struct object_update_result {
3582 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3583 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3584 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3585 /* Hold object_update_results being replied from the remote OUT. */
3586 struct object_update_reply {
3593 /* read update result */
3594 struct out_read_reply {
3601 #define BUT_REQUEST_MAGIC 0xBADE0001
3602 /* Hold batched updates sending to the remote target in a single RPC */
3603 struct batch_update_request {
3604 /* Magic number: BUT_REQUEST_MAGIC. */
3606 /* Number of sub requests packed in this batched RPC: burq_reqmsg[]. */
3608 /* Unused padding field. */
3611 * Sub request message array. As message feild buffers for each sub
3612 * request are packed after padded lustre_msg.lm_buflens[] arrary, thus
3613 * it can locate the next request message via the function
3614 * @batch_update_reqmsg_next() in lustre/include/obj_update.h
3616 struct lustre_msg burq_reqmsg[0];
3619 #define BUT_HEADER_MAGIC 0xBADF0001
3620 /* Header for Batched UpdaTes request */
3621 struct but_update_header {
3622 /* Magic number: BUT_HEADER_MAGIC */
3625 * When the total request buffer length is less than MAX_INLINE_SIZE,
3626 * @buh_count is set with 1 and the batched RPC reqeust can be packed
3628 * Otherwise, @buh_count indicates the IO vector count transferring in
3632 /* inline buffer length when the batched RPC can be packed inline. */
3633 __u32 buh_inline_length;
3634 /* The reply buffer size the client prepared. */
3635 __u32 buh_reply_size;
3636 /* Sub request count in this batched RPC. */
3637 __u32 buh_update_count;
3638 /* Unused padding field. */
3640 /* Inline buffer used when the RPC request can be packed inline. */
3641 __u32 buh_inline_data[0];
3644 struct but_update_buffer {
3649 #define BUT_REPLY_MAGIC 0x00AD0001
3650 /* Batched reply received from a remote targer in a batched RPC. */
3651 struct batch_update_reply {
3652 /* Magic number: BUT_REPLY_MAGIC. */
3654 /* Successful returned sub requests. */
3656 /* Unused padding field. */
3659 * Sub reply message array.
3660 * It can locate the next reply message buffer via the function
3661 * @batch_update_repmsg_next() in lustre/include/obj_update.h
3663 struct lustre_msg burp_repmsg[0];
3667 * Batch update opcode.
3669 enum batch_update_cmd {
3672 BUT_FIRST_OPC = BUT_GETATTR,
3675 /** layout swap request structure
3676 * fid1 and fid2 are in mdt_body
3678 struct mdc_swap_layouts {
3680 } __attribute__((packed));
3682 #define INLINE_RESYNC_ARRAY_SIZE 15
3683 struct close_data_resync_done {
3685 __u32 resync_ids_inline[INLINE_RESYNC_ARRAY_SIZE];
3689 struct lustre_handle cd_handle;
3690 struct lu_fid cd_fid;
3691 __u64 cd_data_version;
3693 __u64 cd_reserved[8];
3694 struct close_data_resync_done cd_resync;
3698 __u32 cd_archive_id;
3699 /* migrate swap layout */
3700 __u64 cd_data_version2;
3704 /* Update llog format */
3706 struct lu_fid uop_fid;
3708 __u16 uop_param_count;
3709 __u16 uop_params_off[];
3710 } __attribute__((packed));
3713 struct update_op uops_op[0];
3716 struct update_params {
3717 struct object_update_param up_params[0];
3720 enum update_records_flag {
3721 UPDATE_RECORD_CONTINUE = 1 >> 0,
3724 * This is the update record format used to store the updates in
3725 * disk. All updates of the operation will be stored in ur_ops.
3726 * All of parameters for updates of the operation will be stored
3728 * To save the space of the record, parameters in ur_ops will only
3729 * remember their offset in ur_params, so to avoid storing duplicate
3730 * parameters in ur_params, which can help us save a lot space for
3731 * operation like creating striped directory.
3733 struct update_records {
3734 __u64 ur_master_transno;
3737 /* If the operation includes multiple updates, then ur_index
3738 * means the index of the update inside the whole updates.
3741 __u32 ur_update_count;
3742 __u32 ur_param_count;
3743 struct update_ops ur_ops;
3744 /* Note ur_ops has a variable size, so comment out
3745 * the following ur_params, in case some use it directly
3746 * update_records->ur_params
3748 * struct update_params ur_params;
3752 struct llog_update_record {
3753 struct llog_rec_hdr lur_hdr;
3754 struct update_records lur_update_rec;
3755 /* Note ur_update_rec has a variable size, so comment out
3756 * the following ur_tail, in case someone use it directly
3758 * struct llog_rec_tail lur_tail;
3762 /* sepol string format is:
3763 * <1-digit for SELinux status>:<policy name>:<policy version>:<policy hash>
3765 /* Max length of the sepol string
3766 * Should be large enough to contain a sha512sum of the policy
3768 #define SELINUX_MODE_LEN 1
3769 #define SELINUX_POLICY_VER_LEN 3 /* 3 chars to leave room for the future */
3770 #define SELINUX_POLICY_HASH_LEN 64
3771 #define LUSTRE_NODEMAP_SEPOL_LENGTH (SELINUX_MODE_LEN + NAME_MAX + \
3772 SELINUX_POLICY_VER_LEN + \
3773 SELINUX_POLICY_HASH_LEN + 3)
3775 /** enums containing the types of ids contained in a nodemap
3776 * kept so other modules (mgs, mdt, etc) can define the type
3779 enum nodemap_id_type {
3785 enum nodemap_tree_type {
3786 NODEMAP_FS_TO_CLIENT,
3787 NODEMAP_CLIENT_TO_FS,
3790 enum nodemap_mapping_modes {
3791 NODEMAP_MAP_BOTH_LEGACY = 0x0, /* for compatibility */
3792 NODEMAP_MAP_UID = 0x01,
3793 NODEMAP_MAP_GID = 0x02,
3794 NODEMAP_MAP_BOTH = 0x03, /* for compatibility */
3795 NODEMAP_MAP_PROJID = 0x04,
3796 NODEMAP_MAP_ALL = NODEMAP_MAP_UID |
3801 enum nodemap_rbac_roles {
3802 NODEMAP_RBAC_FILE_PERMS = 0x00000001,
3803 NODEMAP_RBAC_DNE_OPS = 0x00000002,
3804 NODEMAP_RBAC_QUOTA_OPS = 0x00000004,
3805 NODEMAP_RBAC_BYFID_OPS = 0x00000008,
3806 NODEMAP_RBAC_CHLG_OPS = 0x00000010,
3807 NODEMAP_RBAC_FSCRYPT_ADMIN = 0x00000020,
3808 NODEMAP_RBAC_NONE = (__u32)~(NODEMAP_RBAC_FILE_PERMS |
3809 NODEMAP_RBAC_DNE_OPS |
3810 NODEMAP_RBAC_QUOTA_OPS |
3811 NODEMAP_RBAC_BYFID_OPS |
3812 NODEMAP_RBAC_CHLG_OPS |
3813 NODEMAP_RBAC_FSCRYPT_ADMIN),
3814 NODEMAP_RBAC_ALL = 0xFFFFFFFF, /* future caps ON by default */
3818 * rawobj stuff for GSS
3820 typedef struct netobj_s {
3825 typedef struct rawobj_s {
3832 * following 3 headers must have the same sizes and offsets
3835 __u8 gh_version; /* gss version */
3836 __u8 gh_sp; /* sec part */
3838 __u32 gh_flags; /* wrap flags */
3839 __u32 gh_proc; /* proc */
3840 __u32 gh_seq; /* sequence */
3841 __u32 gh_svc; /* service */
3845 netobj_t gh_handle; /* context handle */
3848 struct gss_rep_header {
3862 struct gss_err_header {
3877 * GSS part of wire context information sent from client, saved and
3878 * used later by server.
3880 struct gss_wire_ctx {
3888 /* This is the lu_ladvise struct which goes out on the wire.
3889 * Corresponds to the userspace arg llapi_lu_ladvise.
3890 * value[1-4] are unspecified fields, used differently by different advices
3893 __u16 lla_advice; /* advice type */
3894 __u16 lla_value1; /* values for different advice types */
3896 __u64 lla_start; /* first byte of extent for advice */
3897 __u64 lla_end; /* last byte of extent for advice */
3902 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3903 * arg llapi_ladvise_hdr.
3904 * value[1-3] are unspecified fields, used differently by different advices
3906 struct ladvise_hdr {
3907 __u32 lah_magic; /* LADVISE_MAGIC */
3908 __u32 lah_count; /* number of advices */
3909 __u64 lah_flags; /* from enum ladvise_flag */
3910 __u32 lah_value1; /* unused */
3911 __u32 lah_value2; /* unused */
3912 __u64 lah_value3; /* unused */
3913 struct lu_ladvise lah_advise[0]; /* advices in this header */
3916 #if defined(__cplusplus)