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 /** \defgroup lu_fid lu_fid
147 extern void lustre_lma_swab(struct lustre_mdt_attrs *lma);
148 extern void lustre_lma_init(struct lustre_mdt_attrs *lma,
149 const struct lu_fid *fid,
150 __u32 compat, __u32 incompat);
151 extern void lustre_loa_swab(struct lustre_ost_attrs *loa,
153 extern void lustre_loa_init(struct lustre_ost_attrs *loa,
154 const struct lu_fid *fid,
155 __u32 compat, __u32 incompat);
157 /* copytool can use any nonnegative integer to represent archive-Ids during
158 * register with MDT thru kuc.
159 * archive num = 0 => all
160 * archive num from 1 to MAX_U32
162 #define LL_HSM_ORIGIN_MAX_ARCHIVE (sizeof(__u32) * 8)
163 /* the max count of archive ids that one agent can support */
164 #define LL_HSM_MAX_ARCHIVES_PER_AGENT 1024
167 * HSM on-disk attributes stored in a separate xattr.
170 /** Bitfield for supported data in this structure. For future use. */
173 /** HSM flags, see hsm_flags enum below */
175 /** backend archive id associated with the file */
177 /** version associated with the last archiving, if any */
180 extern void lustre_hsm_swab(struct hsm_attrs *attrs);
186 /** LASTID file has zero OID */
187 LUSTRE_FID_LASTID_OID = 0UL,
188 /** initial fid id value */
189 LUSTRE_FID_INIT_OID = 1UL
193 * Different FID Format
194 * http://arch.lustre.org/index.php?title=Interoperability_fids_zfs#NEW.0
197 * File IDentifier generated by client from range allocated by the seq service.
198 * First 0x400 sequences [2^33, 2^33 + 0x400] are reserved for system use. Note
199 * that on ldiskfs MDTs that IGIF FIDs can use inode numbers starting at 12,
200 * but this is in the IGIF SEQ rangeand does not conflict with assigned FIDs.
203 * Inode and Generation In FID, a surrogate FID used to globally identify an
204 * existing object on OLD formatted MDT file system. This would only be used on
205 * MDT0 in a DNE filesystem, because there are not expected to be any OLD
206 * formatted DNE filesystems. Belongs to a sequence in [12, 2^32 - 1] range,
207 * where sequence number is inode number, and inode generation is used as OID.
208 * NOTE: This assumes no more than 2^32-1 inodes exist in the MDT filesystem,
209 * which is the maximum possible for an ldiskfs backend. NOTE: This assumes
210 * that the reserved ext3/ext4/ldiskfs inode numbers [0-11] are never visible
211 * to clients, which has always been true.
214 * Object ID in FID, a surrogate FID used to globally identify an existing
215 * object on OLD formatted OST file system. Belongs to a sequence in
216 * [2^32, 2^33 - 1]. Sequence number is calculated as:
217 * 1 << 32 | (ost_index << 16) | ((objid >> 32) & 0xffff)
218 * that is, SEQ consists of 16-bit OST index, and higher 16 bits of object ID.
219 * The generation of unique SEQ values per OST allows the IDIF FIDs to be
220 * identified in the FLD correctly. The OID field is calculated as:
222 * that is, it consists of lower 32 bits of object ID. NOTE This assumes that
223 * no more than 2^48-1 objects have ever been created on an OST, and that no
224 * more than 65535 OSTs are in use. Both are very reasonable assumptions (can
225 * uniquely map all objects on an OST that created 1M objects per second for 9
226 * years, or combinations thereof).
229 * Surrogate FID used to identify an existing object on OLD formatted OST
230 * filesystem. Belongs to the reserved sequence 0, and is used internally prior
231 * to the introduction of FID-on-OST, at which point IDIF will be used to
232 * identify objects as residing on a specific OST.
235 * For Lustre Log objects the object sequence 1 is used. This is compatible with
236 * both OLD and NEW.1 namespaces, as this SEQ number is in the ext3/ldiskfs
237 * reserved inode range and does not conflict with IGIF sequence numbers.
240 * For testing OST IO performance the object sequence 2 is used. This is
241 * compatible with both OLD and NEW.1 namespaces, as this SEQ number is in the
242 * ext3/ldiskfs reserved inode range and does not conflict with IGIF sequence
245 * OST_MDT1 .. OST_MAX:
246 * For testing with multiple MDTs the object sequence 3 through 9 is used,
247 * allowing direct mapping of MDTs 1 through 7 respectively, for a total of 8
248 * MDTs including OST_MDT0. This matches the legacy CMD project "group"
249 * mappings. However, this SEQ range is only for testing prior to any production
250 * DNE release, as the objects in this range conflict across all OSTs, as the
251 * OST index is not part of the FID.
254 * For compatibility with existing OLD OST network protocol structures, the FID
255 * must map onto the o_id and o_gr in a manner that ensures existing objects are
256 * identified consistently for IO, as well as onto the lock namespace to ensure
257 * both IDIFs map onto the same objects for IO as well as resources in the DLM.
260 * resource[] = {o_id, o_seq, 0, 0}; // o_seq == 0 for production releases
262 * DLM NEW.1 FID (this is the same for both the MDT and OST):
263 * resource[] = {SEQ, OID, VER, HASH};
265 * Note that for mapping IDIF values to DLM resource names the o_id may be
266 * larger than the 2^33 reserved sequence numbers for IDIF, so it is possible
267 * for the o_id numbers to overlap FID SEQ numbers in the resource. However, in
268 * all production releases the OLD o_seq field is always zero, and all valid FID
269 * OID values are non-zero, so the lock resources will not collide.
271 * For objects within the IDIF range, group extraction (non-CMD) will be:
272 * o_id = (fid->f_seq & 0x7fff) << 16 | fid->f_oid;
273 * o_seq = 0; // formerly group number
277 * Note that reserved SEQ numbers below 12 will conflict with ldiskfs
278 * inodes in the IGIF namespace, so these reserved SEQ numbers can be
279 * used for other purposes and not risk collisions with existing inodes.
282 FID_SEQ_OST_MDT0 = 0,
283 FID_SEQ_LLOG = 1, /* unnamed llogs */
285 FID_SEQ_UNUSED_START = 3, /* Unused */
286 FID_SEQ_UNUSED_END = 9, /* Unused */
287 FID_SEQ_LLOG_NAME = 10, /* named llogs */
290 FID_SEQ_IGIF_MAX = 0x0ffffffffULL,
291 FID_SEQ_IDIF = 0x100000000ULL,
292 FID_SEQ_IDIF_MAX = 0x1ffffffffULL,
293 /* Normal FID sequence starts from this value, i.e. 1<<33 */
294 FID_SEQ_START = 0x200000000ULL,
295 /* sequence for local pre-defined FIDs listed in local_oid */
296 FID_SEQ_LOCAL_FILE = 0x200000001ULL,
297 FID_SEQ_DOT_LUSTRE = 0x200000002ULL,
298 /* sequence is used for local named objects FIDs generated
299 * by local_object_storage library */
300 FID_SEQ_LOCAL_NAME = 0x200000003ULL,
301 /* Because current FLD will only cache the fid sequence, instead
302 * of oid on the client side, if the FID needs to be exposed to
303 * clients sides, it needs to make sure all of fids under one
304 * sequence will be located in one MDT. */
305 FID_SEQ_SPECIAL = 0x200000004ULL,
306 FID_SEQ_QUOTA = 0x200000005ULL,
307 FID_SEQ_QUOTA_GLB = 0x200000006ULL,
308 FID_SEQ_ROOT = 0x200000007ULL, /* Located on MDT0 */
309 FID_SEQ_LAYOUT_RBTREE = 0x200000008ULL,
310 /* sequence is used for update logs of cross-MDT operation */
311 FID_SEQ_UPDATE_LOG = 0x200000009ULL,
312 /* Sequence is used for the directory under which update logs
314 FID_SEQ_UPDATE_LOG_DIR = 0x20000000aULL,
315 FID_SEQ_NORMAL = 0x200000400ULL,
316 FID_SEQ_LOV_DEFAULT = 0xffffffffffffffffULL
319 #define OBIF_OID_MAX_BITS 32
320 #define OBIF_MAX_OID ((1ULL << OBIF_OID_MAX_BITS) - 1)
321 #define IDIF_OID_MAX_BITS 48
322 #define IDIF_MAX_OID ((1ULL << IDIF_OID_MAX_BITS) - 1)
324 /** OID for FID_SEQ_SPECIAL */
326 /* Big Filesystem Lock to serialize rename operations */
327 FID_OID_SPECIAL_BFL = 1UL,
330 /** OID for FID_SEQ_DOT_LUSTRE */
331 enum dot_lustre_oid {
332 FID_OID_DOT_LUSTRE = 1UL,
333 FID_OID_DOT_LUSTRE_OBF = 2UL,
334 FID_OID_DOT_LUSTRE_LPF = 3UL,
337 /** OID for FID_SEQ_ROOT */
340 FID_OID_ECHO_ROOT = 2UL,
343 struct lu_orphan_rec {
344 /* The MDT-object's FID referenced by the orphan OST-object */
345 struct lu_fid lor_fid;
350 struct lu_orphan_ent {
351 /* The orphan OST-object's FID */
352 struct lu_fid loe_key;
353 struct lu_orphan_rec loe_rec;
356 struct lu_orphan_rec_v2 {
357 struct lu_orphan_rec lor_rec;
358 struct ost_layout lor_layout;
362 struct lu_orphan_ent_v2 {
363 /* The orphan OST-object's FID */
364 struct lu_fid loe_key;
365 struct lu_orphan_rec_v2 loe_rec;
368 struct lu_orphan_rec_v3 {
369 struct lu_orphan_rec lor_rec;
370 struct ost_layout lor_layout;
371 /* The OST-object declared layout version in PFID EA.*/
372 __u32 lor_layout_version;
373 /* The OST-object declared layout range (of version) in PFID EA.*/
379 struct lu_orphan_ent_v3 {
380 /* The orphan OST-object's FID */
381 struct lu_fid loe_key;
382 struct lu_orphan_rec_v3 loe_rec;
387 /** \defgroup lu_dir lu_dir
391 * Enumeration of possible directory entry attributes.
393 * Attributes follow directory entry header in the order they appear in this
396 enum lu_dirent_attrs {
399 LUDA_64BITHASH = 0x0004,
401 /* The following attrs are used for MDT internal only,
402 * not visible to client */
404 /* Something in the record is unknown, to be verified in further. */
405 LUDA_UNKNOWN = 0x0400,
406 /* Ignore this record, go to next directly. */
407 LUDA_IGNORE = 0x0800,
408 /* The system is upgraded, has beed or to be repaired (dryrun). */
409 LUDA_UPGRADE = 0x1000,
410 /* The dirent has been repaired, or to be repaired (dryrun). */
411 LUDA_REPAIR = 0x2000,
412 /* Only check but not repair the dirent inconsistency */
413 LUDA_VERIFY_DRYRUN = 0x4000,
414 /* Verify the dirent consistency */
415 LUDA_VERIFY = 0x8000,
418 #define LU_DIRENT_ATTRS_MASK 0xff00
421 * Layout of readdir pages, as transmitted on wire.
424 /** valid if LUDA_FID is set. */
425 struct lu_fid lde_fid;
426 /** a unique entry identifier: a hash or an offset. */
428 /** total record length, including all attributes. */
432 /** optional variable size attributes following this entry.
433 * taken from enum lu_dirent_attrs.
436 /** name is followed by the attributes indicated in ->ldp_attrs, in
437 * their natural order. After the last attribute, padding bytes are
438 * added to make ->lde_reclen a multiple of 8.
444 * Definitions of optional directory entry attributes formats.
446 * Individual attributes do not have their length encoded in a generic way. It
447 * is assumed that consumer of an attribute knows its format. This means that
448 * it is impossible to skip over an unknown attribute, except by skipping over all
449 * remaining attributes (by using ->lde_reclen), which is not too
450 * constraining, because new server versions will append new attributes at
451 * the end of an entry.
455 * Fid directory attribute: a fid of an object referenced by the entry. This
456 * will be almost always requested by the client and supplied by the server.
458 * Aligned to 8 bytes.
460 /* To have compatibility with 1.8, lets have fid in lu_dirent struct. */
465 * Aligned to 2 bytes.
472 __u64 ldp_hash_start;
476 struct lu_dirent ldp_entries[0];
479 enum lu_dirpage_flags {
481 * dirpage contains no entry.
485 * last entry's lde_hash equals ldp_hash_end.
490 static inline struct lu_dirent *lu_dirent_start(struct lu_dirpage *dp)
492 if (__le32_to_cpu(dp->ldp_flags) & LDF_EMPTY)
495 return dp->ldp_entries;
498 static inline struct lu_dirent *lu_dirent_next(struct lu_dirent *ent)
500 struct lu_dirent *next;
502 if (__le16_to_cpu(ent->lde_reclen) != 0)
503 next = ((void *)ent) + __le16_to_cpu(ent->lde_reclen);
510 static inline __kernel_size_t lu_dirent_calc_size(size_t namelen, __u16 attr)
512 __kernel_size_t size;
514 if (attr & LUDA_TYPE) {
515 const __kernel_size_t align = sizeof(struct luda_type) - 1;
517 size = (sizeof(struct lu_dirent) + namelen + 1 + align) &
519 size += sizeof(struct luda_type);
521 size = sizeof(struct lu_dirent) + namelen + 1;
524 return (size + 7) & ~7;
527 static inline __u16 lu_dirent_type_get(struct lu_dirent *ent)
530 struct luda_type *lt;
533 if (__le32_to_cpu(ent->lde_attrs) & LUDA_TYPE) {
534 const unsigned int align = sizeof(struct luda_type) - 1;
536 len = __le16_to_cpu(ent->lde_namelen);
537 len = (len + align) & ~align;
538 lt = (void *)ent->lde_name + len;
539 type = __le16_to_cpu(lt->lt_type);
545 #define MDS_DIR_END_OFF 0xfffffffffffffffeULL
548 * MDS_READPAGE page size
550 * This is the directory page size packed in MDS_READPAGE RPC.
551 * It's different than PAGE_SIZE because the client needs to
552 * access the struct lu_dirpage header packed at the beginning of
553 * the "page" and without this there isn't any way to know find the
554 * lu_dirpage header is if client and server PAGE_SIZE differ.
556 #define LU_PAGE_SHIFT 12
557 #define LU_PAGE_SIZE (1UL << LU_PAGE_SHIFT)
558 #define LU_PAGE_MASK (~(LU_PAGE_SIZE - 1))
560 #define LU_PAGE_COUNT (1 << (PAGE_SHIFT - LU_PAGE_SHIFT))
564 struct lustre_handle {
567 #define DEAD_HANDLE_MAGIC 0xdeadbeefcafebabeULL
569 static inline bool lustre_handle_is_used(const struct lustre_handle *lh)
571 return lh->cookie != 0;
574 static inline bool lustre_handle_equal(const struct lustre_handle *lh1,
575 const struct lustre_handle *lh2)
577 return lh1->cookie == lh2->cookie;
580 static inline void lustre_handle_copy(struct lustre_handle *tgt,
581 const struct lustre_handle *src)
583 tgt->cookie = src->cookie;
586 /* lustre_msg struct magic. DON'T use swabbed values of MAGIC as magic! */
587 enum lustre_msg_magic {
588 LUSTRE_MSG_MAGIC_V2 = 0x0BD00BD3,
589 LUSTRE_MSG_MAGIC_V2_SWABBED = 0xD30BD00B,
590 LUSTRE_MSG_MAGIC = LUSTRE_MSG_MAGIC_V2
593 /* flags for lm_flags */
595 MSGHDR_AT_SUPPORT = 0x1, /* adaptive timeouts, lm_cksum valid
596 * in early reply messages */
597 MSGHDR_CKSUM_INCOMPAT18 = 0x2, /* compat for 1.8, needs to be set well
598 * beyond 2.8.0 for compatibility */
601 #define lustre_msg lustre_msg_v2
602 /* we depend on this structure to be 8-byte aligned */
603 /* this type is only endian-adjusted in lustre_unpack_msg() */
604 struct lustre_msg_v2 {
605 __u32 lm_bufcount; /* number of buffers in lm_buflens[] */
606 __u32 lm_secflvr; /* 0 = no crypto, or sptlrpc security flavour */
607 __u32 lm_magic; /* RPC version magic = LUSTRE_MSG_MAGIC_V2 */
608 __u32 lm_repsize; /* size of preallocated reply buffer */
609 __u32 lm_cksum; /* CRC32 of ptlrpc_body early reply messages */
610 __u32 lm_flags; /* enum lustre_msghdr MSGHDR_* flags */
611 __u32 lm_opc; /* SUB request opcode in a batch request */
612 __u32 lm_padding_3; /* unused */
613 __u32 lm_buflens[0]; /* length of additional buffers in bytes,
614 * padded to a multiple of 8 bytes. */
616 * message buffers are packed after padded lm_buflens[] array,
617 * padded to a multiple of 8 bytes each to align contents.
621 /* The returned result of the SUB request in a batch request */
622 #define lm_result lm_opc
624 /* ptlrpc_body packet pb_types */
625 #define PTL_RPC_MSG_REQUEST 4711 /* normal RPC request message */
626 #define PTL_RPC_MSG_ERR 4712 /* error reply if request unprocessed */
627 #define PTL_RPC_MSG_REPLY 4713 /* normal RPC reply message */
629 /* ptlrpc_body pb_version ((target_version << 16) | rpc_version) */
630 enum lustre_msg_version {
631 PTLRPC_MSG_VERSION = 0x00000003,
632 LUSTRE_VERSION_MASK = 0xffff0000,
633 LUSTRE_OBD_VERSION = 0x00010000,
634 LUSTRE_MDS_VERSION = 0x00020000,
635 LUSTRE_OST_VERSION = 0x00030000,
636 LUSTRE_DLM_VERSION = 0x00040000,
637 LUSTRE_LOG_VERSION = 0x00050000,
638 LUSTRE_MGS_VERSION = 0x00060000,
641 /* pb_flags that apply to all request messages */
642 /* #define MSG_LAST_REPLAY 0x0001 obsolete 2.0 => {REQ,LOCK}_REPLAY_DONE */
643 #define MSG_RESENT 0x0002 /* was previously sent, no reply seen */
644 #define MSG_REPLAY 0x0004 /* was processed, got reply, recovery */
645 /* #define MSG_AT_SUPPORT 0x0008 obsolete since 1.5, AT always enabled */
646 /* #define MSG_DELAY_REPLAY 0x0010 obsolete since 2.0 */
647 /* #define MSG_VERSION_REPLAY 0x0020 obsolete since 1.8.2, VBR always on */
648 #define MSG_REQ_REPLAY_DONE 0x0040 /* request replay over, locks next */
649 #define MSG_LOCK_REPLAY_DONE 0x0080 /* lock replay over, client done */
651 /* pb_op_flags for connect opcodes: MDS_CONNECT, OST_CONNECT, MGS_CONNECT */
652 #define MSG_CONNECT_RECOVERING 0x00000001 /* target is in recovery */
653 #define MSG_CONNECT_RECONNECT 0x00000002 /* tgt already has client import */
654 #define MSG_CONNECT_REPLAYABLE 0x00000004 /* target supports RPC replay */
655 /* #define MSG_CONNECT_PEER 0x00000008 obsolete since 1.2, removed in 1.5 */
656 #define MSG_CONNECT_LIBCLIENT 0x00000010 /* obsolete since 2.3, removed 2.6 */
657 #define MSG_CONNECT_INITIAL 0x00000020 /* first client connection attempt */
658 /* #define MSG_CONNECT_ASYNC 0x00000040 obsolete since 1.5 */
659 #define MSG_CONNECT_NEXT_VER 0x00000080 /* use next version of lustre_msg */
660 #define MSG_CONNECT_TRANSNO 0x00000100 /* client sent transno in replay */
661 #define MSG_PACK_UID_GID 0x00000200 /* thread UID/GID in ptlrpc_body */
663 /* number of previous object versions in pb_pre_versions[] */
664 #define PTLRPC_NUM_VERSIONS 4
665 /* without gss, ptlrpc_body is put at the first buffer. */
666 struct ptlrpc_body_v3 {
667 struct lustre_handle pb_handle;
668 __u32 pb_type; /* request/reply/err type: PTL_RPC_MSG_* */
669 __u32 pb_version; /* LUSTRE_*_VERSION | PTLRPC_MSG_VERSION */
670 __u32 pb_opc; /* RPC opcodes: MDS_*, OST_*, LDLM_, ... */
671 __u32 pb_status; /* negative Linux x86 error number */
672 __u64 pb_last_xid; /* highest replied XID w/o lower unreplied XID*/
673 __u16 pb_tag; /* multiple modifying RPCs virtual slot index */
676 __u64 pb_last_committed;/* rep: highest pb_transno committed to disk */
677 __u64 pb_transno; /* server-assigned transno for modifying RPCs */
678 __u32 pb_flags; /* req: MSG_* flags */
679 __u32 pb_op_flags; /* req: MSG_CONNECT_* flags */
680 __u32 pb_conn_cnt; /* connect instance of this client on server */
681 __u32 pb_timeout; /* req: max wait time; rep: service estimate */
682 __u32 pb_service_time; /* rep: server arrival to reply in seconds */
683 __u32 pb_limit; /* rep: dynamic DLM LRU lock count limit */
684 __u64 pb_slv; /* rep: dynamic DLM LRU server lock volume */
685 /* VBR: rep: previous pb_version(s) of objects modified by this RPC */
686 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
687 __u64 pb_mbits; /**< match bits for bulk request */
688 /* padding for future needs - fix lustre_swab_ptlrpc_body() also */
689 __u64 pb_padding64_0;
690 __u64 pb_padding64_1;
691 __u32 pb_uid; /* req: process uid, use by tbf rules */
692 __u32 pb_gid; /* req: process gid, use by tbf rules */
693 char pb_jobid[LUSTRE_JOBID_SIZE]; /* req: ASCII jobid from env + NUL */
695 #define ptlrpc_body ptlrpc_body_v3
697 struct ptlrpc_body_v2 {
698 struct lustre_handle pb_handle;
703 __u64 pb_last_xid; /* highest replied XID without lower unreplied XID */
704 __u16 pb_tag; /* virtual slot idx for multiple modifying RPCs */
707 __u64 pb_last_committed;
712 __u32 pb_timeout; /* for req, the deadline, for rep, the service est */
713 __u32 pb_service_time; /* for rep, actual service time, also used for
714 net_latency of req */
717 /* VBR: pre-versions */
718 __u64 pb_pre_versions[PTLRPC_NUM_VERSIONS];
719 __u64 pb_mbits; /**< unused in V2 */
720 /* padding for future needs */
721 __u64 pb_padding64_0;
722 __u64 pb_padding64_1;
723 __u32 pb_uid; /* req: process uid, use by tbf rules */
724 __u32 pb_gid; /* req: process gid, use by tbf rules */
727 /* message body offset for lustre_msg_v2 */
728 /* ptlrpc body offset in all request/reply messages */
729 #define MSG_PTLRPC_BODY_OFF 0
731 /* normal request/reply message record offset */
732 #define REQ_REC_OFF 1
733 #define REPLY_REC_OFF 1
735 /* ldlm request message body offset */
736 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
737 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
739 /* ldlm intent lock message body offset */
740 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
741 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
743 /* ldlm reply message body offset */
744 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
745 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
747 /** only use in req->rq_{req,rep}_swab_mask */
748 #define MSG_PTLRPC_HEADER_OFF 31
751 #define OBD_CONNECT_RDONLY 0x1ULL /* client is read-only */
752 #define OBD_CONNECT_INDEX 0x2ULL /* connect to LOV idx */
753 #define OBD_CONNECT_MDS 0x4ULL /* connect MDT to OST */
754 #define OBD_CONNECT_GRANT 0x8ULL /* fetch grant connect */
755 #define OBD_CONNECT_SRVLOCK 0x10ULL /* server lock for RPC */
756 #define OBD_CONNECT_VERSION 0x20ULL /* versions in OCD */
757 #define OBD_CONNECT_REQPORTAL 0x40ULL /* non-IO portal */
758 #define OBD_CONNECT_ACL 0x80ULL /* access control list */
759 #define OBD_CONNECT_XATTR 0x100ULL /* extended attributes */
760 #define OBD_CONNECT_LARGE_ACL 0x200ULL /* over 32 ACL entries */
761 /* was OBD_CONNECT_TRUNCLOCK 0x400ULL * server locks punch */
762 /* temporary reuse until 2.21.53 to indicate pre-2.15 client, see LU-15478 */
763 #define OBD_CONNECT_OLD_FALLOC 0x400ULL /* no o_valid flags */
764 #define OBD_CONNECT_TRANSNO 0x800ULL /* replay send transno */
765 #define OBD_CONNECT_IBITS 0x1000ULL /* not checked 2.11+ */
766 #define OBD_CONNECT_BARRIER 0x2000ULL /* write barrier */
767 #define OBD_CONNECT_ATTRFID 0x4000ULL /* Server GetAttr FID */
768 #define OBD_CONNECT_NODEVOH 0x8000ULL /* No open handle spec */
769 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Never used, gone 2.9*/
770 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Never used, gone 2.9*/
771 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /* Max bytes per rpc */
772 #define OBD_CONNECT_QUOTA64 0x80000ULL /* Unused since 2.4 */
773 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /* Unused since 2.7 */
774 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /* Unused since 2.7 */
775 #define OBD_CONNECT_CANCELSET 0x400000ULL /* Early batch cancel */
776 #define OBD_CONNECT_SOM 0x800000ULL /* Unused since 2.7 */
777 #define OBD_CONNECT_AT 0x1000000ULL /* client uses AT */
778 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /* LRU resize feature */
779 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /* MDS-MDS connection */
780 #define OBD_CONNECT_REAL 0x8000000ULL /* Unused since 2.8 */
781 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /* Unused since 2.4 */
782 #define OBD_CONNECT_CKSUM 0x20000000ULL /* cksum algo choice */
783 #define OBD_CONNECT_FID 0x40000000ULL /* server handles FIDs */
784 #define OBD_CONNECT_VBR 0x80000000ULL /* version based recov */
785 #define OBD_CONNECT_LOV_V3 0x100000000ULL /* client LOV v3 EA */
786 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* handle grant shrink */
787 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* no orph objid reuse */
788 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* EA size in reply */
789 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
790 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client layout lock */
791 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* 64-bits dir hash */
792 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
793 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recov support */
794 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* ptlrpc_body jobid */
795 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create client umask */
796 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client -EINPROGRESS
797 * RPC error handling */
798 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL /* extra grant params for
799 * space reservation */
800 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* unused since 2.0 */
801 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable LVB type */
802 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosec timestamp */
803 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL /* lightweight connect */
804 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL /* short io */
805 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL /* pings not required */
806 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL /* flk deadlock detect */
807 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL /* create stripe disp */
808 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by FID won't pack
810 #define OBD_CONNECT_LFSCK 0x40000000000000ULL /* allow online LFSCK */
811 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL /* unlink closes file */
812 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* allow multiple change
813 * RPCs in parallel */
814 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
815 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
816 /* was OBD_CONNECT_LOCKAHEAD_OLD 0x1000000000000000ULL old lockahead 2.12-2.13*/
817 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL /* ptlrpc_body matchbit*/
818 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
819 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
820 /* ocd_connect_flags2 flags */
821 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* security context */
822 #define OBD_CONNECT2_LOCKAHEAD 0x2ULL /* ladvise lockahead */
823 #define OBD_CONNECT2_DIR_MIGRATE 0x4ULL /* migrate striped dir */
824 #define OBD_CONNECT2_SUM_STATFS 0x8ULL /* MDT aggregate statfs*/
825 #define OBD_CONNECT2_OVERSTRIPING 0x10ULL /* OST overstriping */
826 #define OBD_CONNECT2_FLR 0x20ULL /* FLR mirror handling */
827 #define OBD_CONNECT2_WBC_INTENTS 0x40ULL /* MDS wb cache intent */
828 #define OBD_CONNECT2_LOCK_CONVERT 0x80ULL /* IBITS lock convert */
829 #define OBD_CONNECT2_ARCHIVE_ID_ARRAY 0x100ULL /* HSM archive_id array*/
830 #define OBD_CONNECT2_INC_XID 0x200ULL /* Increasing xid */
831 #define OBD_CONNECT2_SELINUX_POLICY 0x400ULL /* cli SELinux policy */
832 #define OBD_CONNECT2_LSOM 0x800ULL /* Lazy Size on MDT */
833 #define OBD_CONNECT2_PCC 0x1000ULL /* Persist Client Cache*/
834 #define OBD_CONNECT2_CRUSH 0x2000ULL /* CRUSH dir hash */
835 #define OBD_CONNECT2_ASYNC_DISCARD 0x4000ULL /* async DoM discard */
836 #define OBD_CONNECT2_ENCRYPT 0x8000ULL /* client disk encrypt */
837 #define OBD_CONNECT2_FIDMAP 0x10000ULL /* MDT migrate FID map */
838 #define OBD_CONNECT2_GETATTR_PFID 0x20000ULL /* parent FID getattr */
839 #define OBD_CONNECT2_LSEEK 0x40000ULL /* SEEK_HOLE/DATA RPC */
840 #define OBD_CONNECT2_DOM_LVB 0x80000ULL /* DoM glimpse in LVB */
841 #define OBD_CONNECT2_REP_MBITS 0x100000ULL /* reply mbits, not XID*/
842 #define OBD_CONNECT2_MODE_CONVERT 0x200000ULL /* LDLM mode convert */
843 #define OBD_CONNECT2_BATCH_RPC 0x400000ULL /* Multi-op batch RPCs */
844 #define OBD_CONNECT2_PCCRO 0x800000ULL /* PCC read-only */
845 #define OBD_CONNECT2_MNE_TYPE 0x1000000ULL /* mne_nid_type IPv6 */
846 #define OBD_CONNECT2_LOCK_CONTENTION 0x2000000ULL /* contention detect */
847 #define OBD_CONNECT2_ATOMIC_OPEN_LOCK 0x4000000ULL /* lock on first open */
848 #define OBD_CONNECT2_ENCRYPT_NAME 0x8000000ULL /* name encrypt */
849 #define OBD_CONNECT2_DMV_IMP_INHERIT 0x20000000ULL /* client handle DMV inheritance */
850 #define OBD_CONNECT2_ENCRYPT_FID2PATH 0x40000000ULL /* fid2path enc file */
851 /* For MDS+OSS rolling upgrade interop with 2.16+older, ignored after 2.20.53 */
852 #define OBD_CONNECT2_REPLAY_CREATE 0x80000000ULL /* replay OST_CREATE */
853 #define OBD_CONNECT2_LARGE_NID 0x100000000ULL /* understands large/IPv6 NIDs */
854 #define OBD_CONNECT2_COMPRESS 0x200000000ULL /* compressed file */
855 /* only ZFS servers require a change to support unaligned DIO, so this flag is
856 * ignored for ldiskfs servers */
857 #define OBD_CONNECT2_UNALIGNED_DIO 0x400000000ULL /* unaligned DIO */
858 /* XXX README XXX README XXX README XXX README XXX README XXX README XXX
859 * Please DO NOT add OBD_CONNECT flags before first ensuring that this value
860 * is not in use by some other branch/patch. Email adilger@whamcloud.com
861 * to reserve the new OBD_CONNECT value for use by your feature. Then, submit
862 * a small patch against master and LTS branches that ONLY adds the new flag,
863 * updates obd_connect_names[], adds the flag to check_obd_connect_data(),
864 * and updates wiretests accordingly, so it can be approved and landed easily
865 * to reserve the flag for future use by your feature (submitted separately).
868 #define OCD_HAS_FLAG(ocd, flag) \
869 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flag))
870 #define OCD_HAS_FLAG2(ocd, flag2) (OCD_HAS_FLAG(ocd, FLAGS2) && \
871 !!((ocd)->ocd_connect_flags2 & OBD_CONNECT2_##flag2))
874 #ifdef HAVE_LRU_RESIZE_SUPPORT
875 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
877 #define LRU_RESIZE_CONNECT_FLAG 0
880 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
881 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
882 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
883 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
884 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
885 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
886 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
887 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
888 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
889 OBD_CONNECT_EINPROGRESS | \
890 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
891 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
892 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
893 OBD_CONNECT_FLOCK_DEAD | \
894 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
895 OBD_CONNECT_OPEN_BY_FID | \
896 OBD_CONNECT_DIR_STRIPE | OBD_CONNECT_GRANT | \
897 OBD_CONNECT_SRVLOCK | OBD_CONNECT_BULK_MBITS |\
899 OBD_CONNECT_MULTIMODRPCS |\
900 OBD_CONNECT_SUBTREE | OBD_CONNECT_LARGE_ACL |\
901 OBD_CONNECT_GRANT_PARAM | \
902 OBD_CONNECT_GRANT_SHRINK | \
903 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
905 #define MDT_CONNECT_SUPPORTED2 (OBD_CONNECT2_FILE_SECCTX | \
906 OBD_CONNECT2_DIR_MIGRATE | \
907 OBD_CONNECT2_SUM_STATFS | \
908 OBD_CONNECT2_OVERSTRIPING | \
910 OBD_CONNECT2_LOCK_CONVERT | \
911 OBD_CONNECT2_ARCHIVE_ID_ARRAY | \
912 OBD_CONNECT2_INC_XID | \
913 OBD_CONNECT2_SELINUX_POLICY | \
914 OBD_CONNECT2_LSOM | \
915 OBD_CONNECT2_ASYNC_DISCARD | \
917 OBD_CONNECT2_CRUSH | \
918 OBD_CONNECT2_ENCRYPT | \
919 OBD_CONNECT2_GETATTR_PFID |\
920 OBD_CONNECT2_LSEEK | OBD_CONNECT2_DOM_LVB |\
921 OBD_CONNECT2_REP_MBITS | \
922 OBD_CONNECT2_ATOMIC_OPEN_LOCK | \
923 OBD_CONNECT2_BATCH_RPC | \
924 OBD_CONNECT2_ENCRYPT_NAME | \
925 OBD_CONNECT2_ENCRYPT_FID2PATH | \
926 OBD_CONNECT2_DMV_IMP_INHERIT)
928 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
929 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
930 OBD_CONNECT_INDEX | \
931 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
932 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
933 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
934 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
935 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
936 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
937 OBD_CONNECT_MAX_EASIZE | \
938 OBD_CONNECT_EINPROGRESS | \
939 OBD_CONNECT_JOBSTATS | \
940 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
941 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
942 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
943 OBD_CONNECT_BULK_MBITS | \
944 OBD_CONNECT_GRANT_PARAM | \
945 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
947 #define OST_CONNECT_SUPPORTED2 (OBD_CONNECT2_LOCKAHEAD | OBD_CONNECT2_INC_XID |\
948 OBD_CONNECT2_ENCRYPT | OBD_CONNECT2_LSEEK |\
949 OBD_CONNECT2_REP_MBITS |\
950 OBD_CONNECT2_REPLAY_CREATE)
952 #define ECHO_CONNECT_SUPPORTED (OBD_CONNECT_FID | OBD_CONNECT_FLAGS2)
953 #define ECHO_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS
955 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
956 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
957 OBD_CONNECT_PINGLESS |\
958 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_BARRIER | \
961 #define MGS_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS | \
962 OBD_CONNECT2_LARGE_NID
964 /* Features required for this version of the client to work with server */
965 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_FID | \
966 OBD_CONNECT_ATTRFID | \
969 /* This structure is used for both request and reply.
971 * If we eventually have separate connect data for different types, which we
972 * almost certainly will, then perhaps we stick a union in here. */
973 struct obd_connect_data {
974 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
975 __u32 ocd_version; /* lustre release version number */
976 __u32 ocd_grant; /* initial cache grant amount (bytes) */
977 __u32 ocd_index; /* LOV index to connect to */
978 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
979 __u64 ocd_ibits_known; /* inode bits this client understands */
980 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
981 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
982 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
983 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
984 __u64 ocd_transno; /* first transno from client to be replayed */
985 __u32 ocd_group; /* MDS group on OST */
986 __u32 ocd_cksum_types; /* supported checksum algorithms */
987 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
988 __u32 ocd_instance; /* instance # of this target */
989 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
990 /* Fields after ocd_maxbytes are only accessible by the receiver
991 * if the corresponding flag in ocd_connect_flags is set. Accessing
992 * any field after ocd_maxbytes on the receiver without a valid flag
993 * may result in out-of-bound memory access and kernel oops. */
994 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
995 __u16 padding0; /* READ BELOW! also fix lustre_swab_connect */
996 __u32 padding1; /* READ BELOW! also fix lustre_swab_connect */
997 __u64 ocd_connect_flags2;/* OBD_CONNECT2_* per above */
998 __u64 ocd_compr_type; /* bitmask of supported compression types */
999 __u64 padding4; /* READ BELOW! also fix lustre_swab_connect */
1000 __u64 padding5; /* READ BELOW! also fix lustre_swab_connect */
1001 __u64 padding6; /* READ BELOW! also fix lustre_swab_connect */
1002 __u64 padding7; /* READ BELOW! also fix lustre_swab_connect */
1003 __u64 padding8; /* READ BELOW! also fix lustre_swab_connect */
1004 __u64 padding9; /* READ BELOW! also fix lustre_swab_connect */
1005 __u64 paddingA; /* READ BELOW! also fix lustre_swab_connect */
1006 __u64 paddingB; /* READ BELOW! also fix lustre_swab_connect */
1007 __u64 paddingC; /* READ BELOW! also fix lustre_swab_connect */
1008 __u64 paddingD; /* READ BELOW! also fix lustre_swab_connect */
1009 __u64 paddingE; /* READ BELOW! also fix lustre_swab_connect */
1010 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1012 /* XXX README XXX README XXX README XXX README XXX README XXX README XXX
1013 * Please DO NOT use any fields before first ensuring that this field is
1014 * not in use by some other branch/patch. Email adilger@whamcloud.com to
1015 * reserve the new obd_connect_data field for use by your feature. Then, submit
1016 * a small patch against master and LTS branch that ONLY adds the new field,
1017 * updates lustre_swab_connect(), along with the matching OBD_CONNECT flag,
1018 * and updates wiretests accordingly,so it can be approved and landed easily
1019 * to reserve the field for future use by your feature (submitted separately).
1023 * Supported checksum algorithms. Up to 32 checksum types are supported.
1024 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1025 * Please update DECLARE_CKSUM_NAME in obd_cksum.h when adding a new
1026 * algorithm and also the OBD_FL_CKSUM* flags, OBD_CKSUM_ALL flag,
1027 * OBD_FL_CKSUM_ALL flag and potentially OBD_CKSUM_T10_ALL flag.
1030 OBD_CKSUM_CRC32 = 0x00000001,
1031 OBD_CKSUM_ADLER = 0x00000002,
1032 OBD_CKSUM_CRC32C = 0x00000004,
1033 OBD_CKSUM_RESERVED = 0x00000008,
1034 OBD_CKSUM_T10IP512 = 0x00000010,
1035 OBD_CKSUM_T10IP4K = 0x00000020,
1036 OBD_CKSUM_T10CRC512 = 0x00000040,
1037 OBD_CKSUM_T10CRC4K = 0x00000080,
1040 #define OBD_CKSUM_T10_ALL (OBD_CKSUM_T10IP512 | OBD_CKSUM_T10IP4K | \
1041 OBD_CKSUM_T10CRC512 | OBD_CKSUM_T10CRC4K)
1043 #define OBD_CKSUM_ALL (OBD_CKSUM_CRC32 | OBD_CKSUM_ADLER | OBD_CKSUM_CRC32C | \
1047 * The default checksum algorithm used on top of T10PI GRD tags for RPC.
1048 * Considering that the checksum-of-checksums is only computing CRC32 on a
1049 * 4KB chunk of GRD tags for a 1MB RPC for 512B sectors, or 16KB of GRD
1050 * tags for 16MB of 4KB sectors, this is only 1/256 or 1/1024 of the
1051 * total data being checksummed, so the checksum type used here should not
1052 * affect overall system performance noticeably.
1054 #define OBD_CKSUM_T10_TOP OBD_CKSUM_ADLER
1057 * OST requests: OBDO & OBD request records
1062 OST_REPLY = 0, /* reply ? */
1078 OST_QUOTACHECK = 18, /* not used since 2.4 */
1080 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1084 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
1086 #define OST_FIRST_OPC OST_REPLY
1089 OBD_FL_INLINEDATA = 0x00000001,
1090 OBD_FL_OBDMDEXISTS = 0x00000002,
1091 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1092 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1093 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1094 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1095 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1096 OBD_FL_NO_PRJQUOTA = 0x00000080, /* the object's project is over
1098 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1099 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1100 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1101 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1102 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1103 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1104 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1105 OBD_FL_CKSUM_T10IP512 = 0x00005000, /* T10PI IP cksum, 512B sector */
1106 OBD_FL_CKSUM_T10IP4K = 0x00006000, /* T10PI IP cksum, 4KB sector */
1107 OBD_FL_CKSUM_T10CRC512 = 0x00007000, /* T10PI CRC cksum, 512B sector */
1108 OBD_FL_CKSUM_T10CRC4K = 0x00008000, /* T10PI CRC cksum, 4KB sector */
1109 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1110 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1111 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1112 * XXX: obsoleted - reserved for old
1113 * clients prior than 2.2 */
1114 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1115 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1116 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1117 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1118 OBD_FL_ROOT_SQUASH = 0x00800000, /* root squash */
1119 OBD_FL_ROOT_PRJQUOTA = 0x01000000, /* check prj quota for root */
1120 /* OBD_FL_LOCAL_MASK = 0xF0000000, was local-only flags until 2.10 */
1123 * Note that while the original checksum values were separate bits,
1124 * in 2.x we can actually allow all values from 1-31. T10-PI checksum
1125 * types already use values which are not separate bits.
1127 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1128 OBD_FL_CKSUM_CRC32C | OBD_FL_CKSUM_T10IP512 |
1129 OBD_FL_CKSUM_T10IP4K | OBD_FL_CKSUM_T10CRC512 |
1130 OBD_FL_CKSUM_T10CRC4K,
1132 OBD_FL_NO_QUOTA_ALL = OBD_FL_NO_USRQUOTA | OBD_FL_NO_GRPQUOTA |
1137 * All LOV EA magics should have the same postfix, if some new version
1138 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1139 * Lustre, even though the old version system does not recognizes such
1140 * new magic, it still can distinguish the corrupted cases by checking
1141 * the magic's postfix.
1143 #define LOV_MAGIC_MAGIC 0x0BD0
1144 #define LOV_MAGIC_MASK 0xFFFF
1146 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1147 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1148 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1149 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1150 /* reserved for specifying OSTs */
1151 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1152 #define LOV_MAGIC LOV_MAGIC_V1
1153 #define LOV_MAGIC_COMP_V1 (0x0BD60000 | LOV_MAGIC_MAGIC)
1154 #define LOV_MAGIC_FOREIGN (0x0BD70000 | LOV_MAGIC_MAGIC)
1155 #define LOV_MAGIC_SEL (0x0BD80000 | LOV_MAGIC_MAGIC)
1158 * magic for fully defined striping
1159 * the idea is that we should have different magics for striping "hints"
1160 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1161 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1162 * we can't just change it w/o long way preparation, but we still need a
1163 * mechanism to allow LOD to differentiate hint versus ready striping.
1164 * so, at the moment we do a trick: MDT knows what to expect from request
1165 * depending on the case (replay uses ready striping, non-replay req uses
1166 * hints), so MDT replaces magic with appropriate one and now LOD can
1167 * easily understand what's inside -bzzz
1169 * those *_DEF magics are only used on server side internally, they
1170 * won't be put on wire or disk.
1172 #define LOV_MAGIC_DEFINED 0x10000000
1173 #define LOV_MAGIC_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V1)
1174 #define LOV_MAGIC_V3_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V3)
1175 #define LOV_MAGIC_COMP_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_COMP_V1)
1177 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1178 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1180 #define lov_ost_data lov_ost_data_v1
1181 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1182 struct ost_id l_ost_oi; /* OST object ID */
1183 __u32 l_ost_gen; /* generation of this l_ost_idx */
1184 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1187 #define lov_mds_md lov_mds_md_v1
1188 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1189 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1190 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1191 struct ost_id lmm_oi; /* LOV object ID */
1192 __u32 lmm_stripe_size; /* size of stripe in bytes */
1193 /* lmm_stripe_count used to be __u32 */
1194 __u16 lmm_stripe_count; /* num stripes in use for this object */
1195 __u16 lmm_layout_gen; /* layout generation number */
1196 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1199 #define MAX_MD_SIZE_OLD (sizeof(struct lov_mds_md) + \
1200 4 * sizeof(struct lov_ost_data))
1201 #define MAX_MD_SIZE (sizeof(struct lov_comp_md_v1) + \
1202 4 * (sizeof(struct lov_comp_md_entry_v1) + \
1204 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1206 /* This is the default MDT reply size allocated, should the striping be bigger,
1207 * it will be reallocated in mdt_fix_reply.
1208 * 100 stripes is a bit less than 2.5k of data */
1209 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1210 100 * sizeof(struct lov_ost_data))
1212 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1213 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1214 #define XATTR_USER_PREFIX "user."
1215 #define XATTR_TRUSTED_PREFIX "trusted."
1216 #define XATTR_SECURITY_PREFIX "security."
1217 #define XATTR_ENCRYPTION_PREFIX "encryption."
1219 #define XATTR_NAME_SOM "trusted.som"
1220 #define XATTR_NAME_LOV "trusted.lov"
1221 #define XATTR_NAME_LMA "trusted.lma"
1222 #define XATTR_NAME_LMV "trusted.lmv"
1223 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1224 #define XATTR_NAME_LINK "trusted.link"
1225 #define XATTR_NAME_FID "trusted.fid"
1226 #define XATTR_NAME_VERSION "trusted.version"
1227 #define XATTR_NAME_SOM "trusted.som"
1228 #define XATTR_NAME_HSM "trusted.hsm"
1229 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1230 #define XATTR_NAME_DUMMY "trusted.dummy"
1231 #define XATTR_NAME_PROJID "trusted.projid"
1232 #define XATTR_NAME_DATAVER "trusted.dataver"
1234 #define LL_XATTR_NAME_ENCRYPTION_CONTEXT_OLD XATTR_SECURITY_PREFIX"c"
1235 #define LL_XATTR_NAME_ENCRYPTION_CONTEXT XATTR_ENCRYPTION_PREFIX"c"
1237 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1238 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1240 #define XATTR_NAME_JOB_DEFAULT "user.job"
1241 /* longest allowed jobid xattr name is "user." + 7 chars + null terminator */
1242 #define XATTR_JOB_MAX_LEN 13
1244 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1245 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1246 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1247 struct ost_id lmm_oi; /* LOV object ID */
1248 __u32 lmm_stripe_size; /* size of stripe in bytes */
1249 /* lmm_stripe_count used to be __u32 */
1250 __u16 lmm_stripe_count; /* num stripes in use for this object */
1251 __u16 lmm_layout_gen; /* layout generation number */
1252 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1253 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1256 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1258 if (stripes == LOV_ALL_STRIPES)
1261 if (lmm_magic == LOV_MAGIC_V3)
1262 return sizeof(struct lov_mds_md_v3) +
1263 stripes * sizeof(struct lov_ost_data_v1);
1265 return sizeof(struct lov_mds_md_v1) +
1266 stripes * sizeof(struct lov_ost_data_v1);
1270 lov_mds_md_max_stripe_count(__kernel_size_t buf_size, __u32 lmm_magic)
1272 switch (lmm_magic) {
1273 case LOV_MAGIC_V1: {
1274 struct lov_mds_md_v1 lmm;
1276 if (buf_size < sizeof(lmm))
1279 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1281 case LOV_MAGIC_V3: {
1282 struct lov_mds_md_v3 lmm;
1284 if (buf_size < sizeof(lmm))
1287 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1294 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1295 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1296 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1297 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1298 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1299 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1300 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1301 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1302 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1303 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1304 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1305 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1306 #define OBD_MD_DOM_SIZE (0x00001000ULL) /* Data-on-MDT component size */
1307 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1308 #define OBD_MD_FLPARENT (0x00004000ULL) /* parent FID */
1309 #define OBD_MD_LAYOUT_VERSION (0x00008000ULL) /* OST object layout version */
1310 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1311 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1312 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1313 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1314 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1315 /* OBD_MD_FLQOS (0x00200000ULL) has never been used */
1316 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1317 #define OBD_MD_FLPRJQUOTA (0x00400000ULL) /* over quota flags sent from ost */
1318 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1319 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1320 /* OBD_MD_FLEPOCH (0x04000000ULL) obsolete 2.7.50 */
1321 /* ->mds if epoch opens or closes */
1322 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1323 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1324 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1325 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1326 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1328 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1329 /* OBD_MD_REINT (0x0000000200000000ULL) obsolete 1.8 */
1330 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1331 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1333 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1334 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1335 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1336 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1337 #define OBD_MD_FLAGSTATFS (0x0000010000000000ULL) /* aggregated statfs */
1338 /* OBD_MD_FLMDSCAPA (0x0000020000000000ULL) obsolete 2.7.54 */
1339 /* OBD_MD_FLOSSCAPA (0x0000040000000000ULL) obsolete 2.7.54 */
1340 /* OBD_MD_FLCKSPLIT (0x0000080000000000ULL) obsolete 2.3.58*/
1341 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1342 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1343 * under lock; for xattr
1344 * requests means the
1345 * client holds the lock */
1346 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1348 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1349 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1352 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1353 #define OBD_MD_FLOSTLAYOUT (0x0080000000000000ULL) /* contain ost_layout */
1354 #define OBD_MD_FLPROJID (0x0100000000000000ULL) /* project ID */
1355 #define OBD_MD_SECCTX (0x0200000000000000ULL) /* embed security xattr */
1356 #define OBD_MD_FLLAZYSIZE (0x0400000000000000ULL) /* Lazy size */
1357 #define OBD_MD_FLLAZYBLOCKS (0x0800000000000000ULL) /* Lazy blocks */
1358 #define OBD_MD_FLBTIME (0x1000000000000000ULL) /* birth time */
1359 #define OBD_MD_ENCCTX (0x2000000000000000ULL) /* embed encryption ctx */
1360 #define OBD_MD_NAMEHASH (0x4000000000000000ULL) /* use hash instead of name
1361 * in case of encryption
1364 #define OBD_MD_FLALLQUOTA (OBD_MD_FLUSRQUOTA | \
1365 OBD_MD_FLGRPQUOTA | \
1368 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1369 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1370 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1371 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1372 OBD_MD_FLPARENT | OBD_MD_FLRDEV | OBD_MD_FLGROUP | \
1373 OBD_MD_FLPROJID | OBD_MD_FLBTIME)
1375 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1377 /* don't forget obdo_fid which is way down at the bottom so it can
1378 * come after the definition of llog_cookie */
1382 HSS_CLEARMASK = 0x02,
1383 HSS_ARCHIVE_ID = 0x04,
1386 struct hsm_state_set {
1388 __u32 hss_archive_id;
1390 __u64 hss_clearmask;
1393 /* ost_body.data values for OST_BRW */
1395 #define OBD_BRW_READ 0x01
1396 #define OBD_BRW_WRITE 0x02
1397 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1398 #define OBD_BRW_NDELAY 0x04 /* Non-delay RPC should be issued for
1399 * this page. Non-delay RPCs have bit
1400 * rq_no_delay set. */
1401 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1402 * transfer and is not accounted in
1404 #define OBD_BRW_CHECK 0x10
1405 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1406 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1407 /* OBD_BRW_NOCACHE is currently neither set nor tested */
1408 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1409 #define OBD_BRW_NOQUOTA 0x100 /* do not enforce quota */
1410 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1411 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1412 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1413 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1414 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1415 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1416 * that the client is running low on
1417 * space for unstable pages; asking
1418 * it to sync quickly */
1419 #define OBD_BRW_OVER_PRJQUOTA 0x8000 /* Running out of project quota */
1420 #define OBD_BRW_ROOT_PRJQUOTA 0x10000 /* check project quota for root */
1421 #define OBD_BRW_RDMA_ONLY 0x20000 /* RPC contains RDMA-only pages*/
1422 #define OBD_BRW_SYS_RESOURCE 0x40000 /* page has CAP_SYS_RESOURCE */
1423 #define OBD_BRW_COMPRESSED 0x80000 /* data compressed on client */
1425 #define OBD_BRW_OVER_ALLQUOTA (OBD_BRW_OVER_USRQUOTA | \
1426 OBD_BRW_OVER_GRPQUOTA | \
1427 OBD_BRW_OVER_PRJQUOTA)
1429 #define OBD_BRW_DONE 0x40000000UL /*
1430 * osd-ldiskfs inernal,
1431 * IO has been issued before
1433 #define OBD_BRW_LOCAL1 0x80000000UL /*
1434 * osd-ldiskfs internal,
1435 * page mapped to real block
1438 #define OBD_BRW_LOCALS (OBD_BRW_LOCAL1 | OBD_BRW_DONE)
1440 #define OBD_MAX_GRANT 0x7fffffffUL /* Max grant allowed to one client: 2 GiB */
1442 #define OBD_OBJECT_EOF LUSTRE_EOF
1444 #define OST_MIN_PRECREATE 32
1445 #define OST_MAX_PRECREATE 20000
1448 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1449 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1450 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1451 * high 16 bits in 2.4 and later */
1452 __u32 ioo_bufcnt; /* number of niobufs for this object */
1455 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1456 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1457 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. */
1458 #define IOOBJ_MAX_BRW_BITS 16
1459 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1460 #define ioobj_max_brw_set(ioo, num) \
1461 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1463 /* multiple of 8 bytes => can array */
1464 struct niobuf_remote {
1470 /* lock value block communicated between the filter and llite */
1472 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1473 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1474 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1475 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1476 #define OST_LVB_IS_ERR(blocks) \
1477 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1478 #define OST_LVB_SET_ERR(blocks, rc) \
1479 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1480 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1503 * lquota data structures
1506 /* The lquota_id structure is an union of all the possible identifier types that
1507 * can be used with quota, this includes:
1510 * - a FID which can be used for per-directory quota in the future */
1512 struct lu_fid qid_fid; /* FID for per-directory quota */
1513 __u64 qid_uid; /* user identifier */
1514 __u64 qid_gid; /* group identifier */
1515 __u64 qid_projid; /* project identifier */
1518 /* quotactl management */
1519 struct obd_quotactl {
1521 __u32 qc_type; /* see Q_* flag below */
1524 struct obd_dqinfo qc_dqinfo;
1525 struct obd_dqblk qc_dqblk;
1529 #define Q_COPY(out, in, member) (out)->member = (in)->member
1532 * - in and out maybe a type of struct if_quotactl or struct obd_quotactl
1533 * - in and out need not be of the same type.
1535 #define __QCTL_COPY(out, in, need_pname) \
1537 Q_COPY(out, in, qc_cmd); \
1538 Q_COPY(out, in, qc_type); \
1539 Q_COPY(out, in, qc_id); \
1540 Q_COPY(out, in, qc_stat); \
1541 Q_COPY(out, in, qc_dqinfo); \
1542 Q_COPY(out, in, qc_dqblk); \
1543 if (need_pname && LUSTRE_Q_CMD_IS_POOL(in->qc_cmd)) { \
1544 size_t len = strnlen(in->qc_poolname, LOV_MAXPOOLNAME); \
1546 memcpy(out->qc_poolname, in->qc_poolname, len); \
1547 out->qc_poolname[len] = '\0'; \
1551 #define QCTL_COPY(out, in) __QCTL_COPY(out, in, true)
1552 #define QCTL_COPY_NO_PNAME(out, in) __QCTL_COPY(out, in, false)
1554 /* Body of quota request used for quota acquire/release RPCs between quota
1555 * master (aka QMT) and slaves (ak QSD). */
1557 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1558 * and type (data or metadata) as well as
1559 * the quota type (user or group). */
1560 union lquota_id qb_id; /* uid or gid or directory FID */
1561 __u32 qb_flags; /* see below */
1563 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1564 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1565 __u64 qb_slv_ver; /* slave index file version */
1566 struct lustre_handle qb_lockh; /* per-ID lock handle */
1567 struct lustre_handle qb_glb_lockh; /* global lock handle */
1568 __u64 qb_padding1[4];
1571 /* When the quota_body is used in the reply of quota global intent
1572 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1573 #define qb_slv_fid qb_fid
1574 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1576 #define qb_qunit qb_usage
1578 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1579 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1580 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1581 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1583 /* Quota types currently supported */
1585 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1586 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1587 LQUOTA_TYPE_PRJ = 0x02, /* maps to PRJQUOTA */
1591 /* There are 2 different resource types on which a quota limit can be enforced:
1592 * - inodes on the MDTs
1593 * - blocks on the OSTs */
1595 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1596 LQUOTA_RES_DT = 0x02,
1598 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1600 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1603 * Space accounting support
1604 * Format of an accounting record, providing disk usage information for a given
1607 struct lquota_acct_rec { /* 16 bytes */
1608 __u64 bspace; /* current space in use */
1609 __u64 ispace; /* current # inodes in use */
1613 * Global quota index support
1614 * Format of a global record, providing global quota settings for a given quota
1617 struct lquota_glb_rec { /* 32 bytes */
1618 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1619 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1620 __u64 qbr_time; /* grace time, in seconds */
1621 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1626 * Slave index support
1627 * Format of a slave record, recording how much space is granted to a given
1630 struct lquota_slv_rec { /* 8 bytes */
1631 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1632 * in #inodes or kbytes */
1635 /* Data structures associated with the quota locks */
1637 /* Glimpse descriptor used for the index & per-ID quota locks */
1638 struct ldlm_gl_lquota_desc {
1639 union lquota_id gl_id; /* quota ID subject to the glimpse */
1640 __u64 gl_flags; /* see LQUOTA_FL* below */
1641 __u64 gl_ver; /* new index version */
1642 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1643 __u64 gl_softlimit; /* new softlimit */
1647 #define gl_qunit gl_hardlimit /* current qunit value used when
1648 * glimpsing per-ID quota locks */
1650 /* quota glimpse flags */
1651 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1653 /* LVB used with quota (global and per-ID) locks */
1655 __u64 lvb_flags; /* see LQUOTA_FL* above */
1656 __u64 lvb_id_may_rel; /* space that might be released later */
1657 __u64 lvb_id_rel; /* space released by the slave for this ID */
1658 __u64 lvb_id_qunit; /* current qunit value */
1662 /* LVB used with global quota lock */
1663 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1671 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1680 MDS_GETATTR_NAME = 34,
1685 MDS_DISCONNECT = 39,
1688 MDS_PIN = 42, /* obsolete, never used in a release */
1689 MDS_UNPIN = 43, /* obsolete, never used in a release */
1691 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1693 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1696 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1698 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1700 MDS_HSM_STATE_GET = 54,
1701 MDS_HSM_STATE_SET = 55,
1702 MDS_HSM_ACTION = 56,
1703 MDS_HSM_PROGRESS = 57,
1704 MDS_HSM_REQUEST = 58,
1705 MDS_HSM_CT_REGISTER = 59,
1706 MDS_HSM_CT_UNREGISTER = 60,
1707 MDS_SWAP_LAYOUTS = 61,
1713 #define MDS_FIRST_OPC MDS_GETATTR
1716 /* opcodes for object update */
1722 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1742 /* the disposition of the intent outlines what was executed */
1743 #define DISP_IT_EXECD 0x00000001
1744 #define DISP_LOOKUP_EXECD 0x00000002
1745 #define DISP_LOOKUP_NEG 0x00000004
1746 #define DISP_LOOKUP_POS 0x00000008
1747 #define DISP_OPEN_CREATE 0x00000010
1748 #define DISP_OPEN_OPEN 0x00000020
1749 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1750 #define DISP_ENQ_OPEN_REF 0x00800000
1751 #define DISP_ENQ_CREATE_REF 0x01000000
1752 #define DISP_OPEN_LOCK 0x02000000
1753 #define DISP_OPEN_LEASE 0x04000000
1754 #define DISP_OPEN_STRIPE 0x08000000
1755 #define DISP_OPEN_DENY 0x10000000
1757 /* INODE LOCK PARTS */
1758 enum mds_ibits_locks {
1759 MDS_INODELOCK_LOOKUP = 0x000001, /* For namespace, dentry etc. Was
1760 * used to protect permission (mode,
1761 * owner, group, etc) before 2.4. */
1762 MDS_INODELOCK_UPDATE = 0x000002, /* size, links, timestamps */
1763 MDS_INODELOCK_OPEN = 0x000004, /* For opened files */
1764 MDS_INODELOCK_LAYOUT = 0x000008, /* for layout */
1766 /* The PERM bit is added in 2.4, and is used to protect permission
1767 * (mode, owner, group, ACL, etc.) separate from LOOKUP lock.
1768 * For remote directories (in DNE) these locks will be granted by
1769 * different MDTs (different LDLM namespace).
1771 * For local directory, the MDT always grants UPDATE|PERM together.
1772 * For remote directory, master MDT (where remote directory is) grants
1773 * UPDATE|PERM, and remote MDT (where name entry is) grants LOOKUP_LOCK.
1775 MDS_INODELOCK_PERM = 0x000010,
1776 MDS_INODELOCK_XATTR = 0x000020, /* non-permission extended attrs */
1777 MDS_INODELOCK_DOM = 0x000040, /* Data for Data-on-MDT files */
1778 /* Do not forget to increase MDS_INODELOCK_NUMBITS when adding bits */
1780 #define MDS_INODELOCK_NUMBITS 7
1781 /* This FULL lock is useful to take on unlink sort of operations */
1782 #define MDS_INODELOCK_FULL ((1 << MDS_INODELOCK_NUMBITS) - 1)
1783 /* DOM lock shouldn't be canceled early, use this macro for ELC */
1784 #define MDS_INODELOCK_ELC (MDS_INODELOCK_FULL & ~MDS_INODELOCK_DOM)
1786 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1787 * but was moved into name[1] along with the OID to avoid consuming the
1788 * name[2,3] fields that need to be used for the quota id (also a FID). */
1790 LUSTRE_RES_ID_SEQ_OFF = 0,
1791 LUSTRE_RES_ID_VER_OID_OFF = 1,
1792 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1793 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1794 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1795 LUSTRE_RES_ID_HSH_OFF = 3
1798 #define MDS_STATUS_CONN 1
1799 #define MDS_STATUS_LOV 2
1802 /* these should be identical to their EXT4_*_FL counterparts, they are
1803 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1804 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1805 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1806 LUSTRE_APPEND_FL = 0x00000020, /* file writes may only append */
1807 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1808 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1809 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1810 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1811 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1812 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1813 LUSTRE_PROJINHERIT_FL = 0x20000000, /* Create with parents projid */
1815 /* These flags will not be identical to any EXT4_*_FL counterparts,
1816 * and only reserved for lustre purpose. Note: these flags might
1817 * be conflict with some of EXT4 flags, so
1818 * 1. these conflict flags needs to be removed when the flag is
1819 * wired by la_flags see osd_attr_get().
1820 * 2. If these flags needs to be stored into inode, they will be
1821 * stored in LMA. see LMAI_XXXX */
1822 LUSTRE_ORPHAN_FL = 0x00002000,
1823 LUSTRE_SET_SYNC_FL = 0x00040000, /* Synchronous setattr on OSTs */
1824 LUSTRE_ENCRYPT_FL = 0x00800000, /* encrypted file */
1826 LUSTRE_LMA_FL_MASKS = LUSTRE_ENCRYPT_FL | LUSTRE_ORPHAN_FL,
1829 #ifndef FS_XFLAG_SYNC
1830 #define FS_XFLAG_SYNC 0x00000020 /* all writes synchronous */
1832 #ifndef FS_XFLAG_NOATIME
1833 #define FS_XFLAG_NOATIME 0x00000040 /* do not update access time */
1835 #ifndef FS_XFLAG_IMMUTABLE
1836 #define FS_XFLAG_IMMUTABLE 0x00000008 /* file cannot be modified */
1838 #ifndef FS_XFLAG_APPEND
1839 #define FS_XFLAG_APPEND 0x00000010 /* all writes append */
1841 #ifndef FS_XFLAG_PROJINHERIT
1842 #define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
1845 /* 64 possible states */
1846 enum md_transient_state {
1847 MS_RESTORE = (1 << 0), /* restore is running */
1851 struct lu_fid mbo_fid1;
1852 struct lu_fid mbo_fid2;
1853 struct lustre_handle mbo_open_handle;
1855 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1859 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1860 __u64 mbo_version; /* was mbo_ioepoch before 2.11 */
1861 __u64 mbo_t_state; /* transient file state defined in
1862 * enum md_transient_state
1863 * was "ino" until 2.4.0 */
1866 __u32 mbo_capability;
1870 __u32 mbo_flags; /* most replies: LUSTRE_*_FL file attributes,
1871 * data_version: OBD_FL_* flags
1874 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1875 __u32 mbo_layout_gen; /* was "generation" until 2.4.0 */
1877 __u32 mbo_eadatasize;
1879 __u32 mbo_max_mdsize;
1880 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1881 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1882 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1884 __u64 mbo_dom_size; /* size of DOM component */
1885 __u64 mbo_dom_blocks; /* blocks consumed by DOM component */
1887 __u64 mbo_padding_9; /* also fix lustre_swab_mdt_body */
1888 __u64 mbo_padding_10;
1891 struct mdt_ioepoch {
1892 struct lustre_handle mio_open_handle;
1893 __u64 mio_unused1; /* was ioepoch */
1894 __u32 mio_unused2; /* was flags */
1898 /* permissions for md_perm.mp_perm */
1900 CFS_SETUID_PERM = 0x01,
1901 CFS_SETGID_PERM = 0x02,
1902 CFS_SETGRP_PERM = 0x04,
1905 struct mdt_rec_setattr {
1915 __u32 sa_padding_1_h;
1916 struct lu_fid sa_fid;
1925 __u32 sa_attr_flags;
1927 __u32 sa_bias; /* some operation flags */
1934 * Attribute flags used in mdt_rec_setattr::sa_valid.
1935 * The kernel's #defines for ATTR_* should not be used over the network
1936 * since the client and MDS may run different kernels (see bug 13828)
1937 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1939 enum mds_attr_flags {
1940 MDS_ATTR_MODE = 0x1ULL, /* = 1 */
1941 MDS_ATTR_UID = 0x2ULL, /* = 2 */
1942 MDS_ATTR_GID = 0x4ULL, /* = 4 */
1943 MDS_ATTR_SIZE = 0x8ULL, /* = 8 */
1944 MDS_ATTR_ATIME = 0x10ULL, /* = 16 */
1945 MDS_ATTR_MTIME = 0x20ULL, /* = 32 */
1946 MDS_ATTR_CTIME = 0x40ULL, /* = 64 */
1947 MDS_ATTR_ATIME_SET = 0x80ULL, /* = 128 */
1948 MDS_ATTR_MTIME_SET = 0x100ULL, /* = 256 */
1949 MDS_ATTR_FORCE = 0x200ULL, /* = 512, change it */
1950 MDS_ATTR_ATTR_FLAG = 0x400ULL, /* = 1024 */
1951 MDS_ATTR_KILL_SUID = 0x800ULL, /* = 2048 */
1952 MDS_ATTR_KILL_SGID = 0x1000ULL, /* = 4096 */
1953 MDS_ATTR_CTIME_SET = 0x2000ULL, /* = 8192 */
1954 MDS_ATTR_FROM_OPEN = 0x4000ULL, /* = 16384, from open O_TRUNC */
1955 MDS_ATTR_BLOCKS = 0x8000ULL, /* = 32768 */
1956 MDS_ATTR_PROJID = 0x10000ULL, /* = 65536 */
1957 MDS_ATTR_LSIZE = 0x20000ULL, /* = 131072 */
1958 MDS_ATTR_LBLOCKS = 0x40000ULL, /* = 262144 */
1959 MDS_ATTR_OVERRIDE = 0x2000000ULL, /* = 33554432 */
1963 /* MDS_CHECK_SPLIT = 1 << 0, obsolete before 2.3.58 */
1964 /* used for remote object getattr/open by name: in the original
1965 * getattr/open request, MDT found the object against name is on another
1966 * MDT, then packed FID and LOOKUP lock in reply and returned -EREMOTE,
1967 * and client knew it's a remote object, then set this flag in
1968 * getattr/open request and sent to the corresponding MDT to finish
1969 * getattr/open, which fetched attributes and UPDATE lock/opened file.
1971 MDS_CROSS_REF = 1 << 1,
1972 /* MDS_VTX_BYPASS = 1 << 2, obsolete since 2.3.54 */
1973 MDS_PERM_BYPASS = 1 << 3,
1974 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1975 MDS_QUOTA_IGNORE = 1 << 5,
1976 /* MDS_CLOSE_CLEANUP = 1 << 6, obsolete since 2.3.51 */
1977 MDS_KEEP_ORPHAN = 1 << 7,
1978 MDS_RECOV_OPEN = 1 << 8,
1979 MDS_DATA_MODIFIED = 1 << 9,
1980 MDS_CREATE_VOLATILE = 1 << 10,
1981 MDS_OWNEROVERRIDE = 1 << 11,
1982 MDS_HSM_RELEASE = 1 << 12,
1983 MDS_CLOSE_MIGRATE = 1 << 13,
1984 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1985 MDS_CLOSE_LAYOUT_MERGE = 1 << 15,
1986 MDS_CLOSE_RESYNC_DONE = 1 << 16,
1987 MDS_CLOSE_LAYOUT_SPLIT = 1 << 17,
1988 MDS_TRUNC_KEEP_LEASE = 1 << 18,
1989 MDS_PCC_ATTACH = 1 << 19,
1990 MDS_CLOSE_UPDATE_TIMES = 1 << 20,
1991 /* setstripe create only, don't restripe if target exists */
1992 MDS_SETSTRIPE_CREATE = 1 << 21,
1993 MDS_FID_OP = 1 << 22,
1994 /* migrate dirent only */
1995 MDS_MIGRATE_NSONLY = 1 << 23,
1996 /* create with default LMV from client */
1997 MDS_CREATE_DEFAULT_LMV = 1 << 24,
2000 #define MDS_CLOSE_INTENT (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP | \
2001 MDS_CLOSE_LAYOUT_MERGE | MDS_CLOSE_LAYOUT_SPLIT | \
2002 MDS_CLOSE_RESYNC_DONE)
2004 /* instance of mdt_reint_rec */
2005 struct mdt_rec_create {
2013 __u32 cr_suppgid1_h;
2015 __u32 cr_suppgid2_h;
2016 struct lu_fid cr_fid1;
2017 struct lu_fid cr_fid2;
2018 struct lustre_handle cr_open_handle_old; /* in case of open replay */
2022 __u32 cr_archive_id;
2025 __u64 cr_padding_1; /* rr_blocks */
2028 /* use of helpers set/get_mrc_cr_flags() is needed to access
2029 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2030 * extend cr_flags size without breaking 1.8 compat */
2031 __u32 cr_flags_l; /* for use with open, low 32 bits */
2032 __u32 cr_flags_h; /* for use with open, high 32 bits */
2033 __u32 cr_umask; /* umask for create */
2034 __u32 cr_padding_4; /* rr_padding_4 */
2037 /* instance of mdt_reint_rec */
2038 struct mdt_rec_link {
2046 __u32 lk_suppgid1_h;
2048 __u32 lk_suppgid2_h;
2049 struct lu_fid lk_fid1;
2050 struct lu_fid lk_fid2;
2052 __u64 lk_padding_1; /* rr_atime */
2053 __u64 lk_padding_2; /* rr_ctime */
2054 __u64 lk_padding_3; /* rr_size */
2055 __u64 lk_padding_4; /* rr_blocks */
2057 __u32 lk_padding_5; /* rr_mode */
2058 __u32 lk_padding_6; /* rr_flags */
2059 __u32 lk_padding_7; /* rr_padding_2 */
2060 __u32 lk_padding_8; /* rr_padding_3 */
2061 __u32 lk_padding_9; /* rr_padding_4 */
2064 /* instance of mdt_reint_rec */
2065 struct mdt_rec_unlink {
2073 __u32 ul_suppgid1_h;
2075 __u32 ul_suppgid2_h;
2076 struct lu_fid ul_fid1;
2077 struct lu_fid ul_fid2;
2079 __u64 ul_padding_2; /* rr_atime */
2080 __u64 ul_padding_3; /* rr_ctime */
2081 __u64 ul_padding_4; /* rr_size */
2082 __u64 ul_padding_5; /* rr_blocks */
2085 __u32 ul_padding_6; /* rr_flags */
2086 __u32 ul_padding_7; /* rr_padding_2 */
2087 __u32 ul_padding_8; /* rr_padding_3 */
2088 __u32 ul_padding_9; /* rr_padding_4 */
2091 /* instance of mdt_reint_rec */
2092 struct mdt_rec_rename {
2100 __u32 rn_suppgid1_h;
2102 __u32 rn_suppgid2_h;
2103 struct lu_fid rn_fid1;
2104 struct lu_fid rn_fid2;
2106 __u64 rn_padding_1; /* rr_atime */
2107 __u64 rn_padding_2; /* rr_ctime */
2108 __u64 rn_padding_3; /* rr_size */
2109 __u64 rn_padding_4; /* rr_blocks */
2110 __u32 rn_bias; /* some operation flags */
2111 __u32 rn_mode; /* cross-ref rename has mode */
2112 __u32 rn_padding_5; /* rr_flags */
2113 __u32 rn_padding_6; /* rr_padding_2 */
2114 __u32 rn_padding_7; /* rr_padding_3 */
2115 __u32 rn_padding_8; /* rr_padding_4 */
2118 /* instance of mdt_reint_rec */
2119 struct mdt_rec_setxattr {
2127 __u32 sx_suppgid1_h;
2129 __u32 sx_suppgid2_h;
2130 struct lu_fid sx_fid;
2131 __u64 sx_padding_1; /* These three are rr_fid2 */
2136 __u64 sx_padding_5; /* rr_ctime */
2137 __u64 sx_padding_6; /* rr_size */
2138 __u64 sx_padding_7; /* rr_blocks */
2141 __u32 sx_padding_8; /* rr_flags */
2142 __u32 sx_padding_9; /* rr_padding_2 */
2143 __u32 sx_padding_10; /* rr_padding_3 */
2144 __u32 sx_padding_11; /* rr_padding_4 */
2147 /* instance of mdt_reint_rec
2148 * FLR: for file resync MDS_REINT_RESYNC RPC. */
2149 struct mdt_rec_resync {
2157 __u32 rs_suppgid1_h;
2159 __u32 rs_suppgid2_h;
2160 struct lu_fid rs_fid;
2161 __u8 rs_padding0[sizeof(struct lu_fid)];
2162 struct lustre_handle rs_lease_handle; /* rr_mtime */
2163 __s64 rs_padding1; /* rr_atime */
2164 __s64 rs_padding2; /* rr_ctime */
2165 __u64 rs_padding3; /* rr_size */
2166 __u64 rs_padding4; /* rr_blocks */
2168 __u32 rs_padding5; /* rr_mode */
2169 __u32 rs_padding6; /* rr_flags */
2170 __u32 rs_padding7; /* rr_flags_h */
2171 __u32 rs_padding8; /* rr_umask */
2173 __u16 rs_padding9; /* rr_padding_4 */
2177 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2178 * Do NOT change the size of various members, otherwise the value
2179 * will be broken in lustre_swab_mdt_rec_reint().
2181 * If you add new members in other mdt_reint_xxx structres and need to use the
2182 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2184 struct mdt_rec_reint {
2192 __u32 rr_suppgid1_h;
2194 __u32 rr_suppgid2_h;
2195 struct lu_fid rr_fid1;
2196 struct lu_fid rr_fid2;
2208 __u16 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2211 #define LMV_DESC_QOS_MAXAGE_DEFAULT 60 /* Seconds */
2213 /* lmv structures */
2215 __u32 ld_tgt_count; /* how many MDS's */
2216 __u32 ld_active_tgt_count; /* how many active */
2217 __u32 ld_default_stripe_count; /* how many objects are used */
2218 __u32 ld_pattern; /* default hash pattern */
2219 __u64 ld_default_hash_size;
2220 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2221 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2222 __u32 ld_qos_maxage; /* in second */
2223 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2224 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2225 struct obd_uuid ld_uuid;
2228 /* LMV layout EA, and it will be stored both in master and slave object */
2229 struct lmv_mds_md_v1 {
2231 __u32 lmv_stripe_count;
2232 __u32 lmv_master_mdt_index; /* On master object, it is master
2233 * MDT index, on slave object, it
2234 * is stripe index of the slave obj */
2235 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2236 * which hash function to be used,
2237 * Note: only lower 16 bits is being
2238 * used for now. Higher 16 bits will
2239 * be used to mark the object status,
2240 * for example migrating or dead. */
2241 __u32 lmv_layout_version; /* increased each time layout changed,
2242 * by directory migration, restripe
2244 __u32 lmv_migrate_offset; /* once this is set, it means this
2245 * directory is been migrated, stripes
2246 * before this offset belong to target,
2247 * from this to source. */
2248 __u32 lmv_migrate_hash; /* hash type of source stripes of
2249 * migrating directory */
2252 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
2253 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2256 /* stripe count before directory split */
2257 #define lmv_split_offset lmv_migrate_offset
2258 /* stripe count after directory merge */
2259 #define lmv_merge_offset lmv_migrate_offset
2260 /* directory hash type after merge */
2261 #define lmv_merge_hash lmv_migrate_hash
2263 /* foreign LMV EA */
2264 struct lmv_foreign_md {
2265 __u32 lfm_magic; /* magic number = LMV_MAGIC_FOREIGN */
2266 __u32 lfm_length; /* length of lfm_value */
2267 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
2268 __u32 lfm_flags; /* flags, type specific */
2269 char lfm_value[]; /* free format value */
2272 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2273 #define LMV_MAGIC LMV_MAGIC_V1
2275 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2276 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2277 #define LMV_MAGIC_FOREIGN 0x0CD50CD0 /* magic for lmv foreign */
2280 * The FNV-1a hash algorithm is as follows:
2281 * hash = FNV_offset_basis
2282 * for each octet_of_data to be hashed
2283 * hash = hash XOR octet_of_data
2284 * hash = hash × FNV_prime
2286 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2288 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2289 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2291 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2292 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2293 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, __kernel_size_t size)
2295 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2296 const unsigned char *p = buf;
2299 for (i = 0; i < size; i++) {
2301 hash *= LUSTRE_FNV_1A_64_PRIME;
2307 /* CRUSH placement group count */
2308 #define LMV_CRUSH_PG_COUNT 4096
2312 struct lmv_mds_md_v1 lmv_md_v1;
2313 struct lmv_user_md lmv_user_md;
2314 struct lmv_foreign_md lmv_foreign_md;
2317 static inline __kernel_ssize_t lmv_mds_md_size(int stripe_count,
2318 unsigned int lmm_magic)
2320 __kernel_ssize_t len = -EINVAL;
2322 switch (lmm_magic) {
2323 case LMV_MAGIC_V1: {
2324 struct lmv_mds_md_v1 *lmm1;
2326 len = sizeof(*lmm1);
2327 len += stripe_count * sizeof(lmm1->lmv_stripe_fids[0]);
2335 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2337 switch (__le32_to_cpu(lmm->lmv_magic)) {
2339 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2340 case LMV_USER_MAGIC:
2341 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2347 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2349 switch (__le32_to_cpu(lmm->lmv_magic)) {
2351 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2352 case LMV_USER_MAGIC:
2353 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2363 FLD_FIRST_OPC = FLD_QUERY
2369 SEQ_FIRST_OPC = SEQ_QUERY
2373 SEQ_ALLOC_SUPER = 0,
2385 LFSCK_NOTIFY = 1101,
2388 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2392 * LOV data structures
2395 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2396 /* The size of the buffer the lov/mdc reserves for the
2397 * array of UUIDs returned by the MDS. With the current
2398 * protocol, this will limit the max number of OSTs per LOV */
2400 #define LOV_DESC_MAGIC 0xB0CCDE5C
2401 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2402 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2404 /* LOV settings descriptor (should only contain static info) */
2406 __u32 ld_tgt_count; /* how many OBD's */
2407 __u32 ld_active_tgt_count; /* how many active */
2408 __s32 ld_default_stripe_count; /* how many objects are used */
2409 __u32 ld_pattern; /* default PATTERN_RAID0 */
2410 __u64 ld_default_stripe_size; /* in bytes */
2411 __s64 ld_default_stripe_offset; /* starting OST index */
2412 __u32 ld_padding_0; /* unused */
2413 __u32 ld_qos_maxage; /* in second */
2414 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2415 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2416 struct obd_uuid ld_uuid;
2419 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2424 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2429 LDLM_BL_CALLBACK = 104,
2430 LDLM_CP_CALLBACK = 105,
2431 LDLM_GL_CALLBACK = 106,
2432 LDLM_SET_INFO = 107,
2435 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2437 #define RES_NAME_SIZE 4
2438 struct ldlm_res_id {
2439 __u64 name[RES_NAME_SIZE];
2442 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2443 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2444 (unsigned long long)(res)->lr_name.name[1], \
2445 (unsigned long long)(res)->lr_name.name[2], \
2446 (unsigned long long)(res)->lr_name.name[3]
2463 #define LCK_MODE_NUM 9
2473 #define LDLM_MIN_TYPE LDLM_PLAIN
2475 struct ldlm_extent {
2481 static inline bool ldlm_extent_equal(const struct ldlm_extent *ex1,
2482 const struct ldlm_extent *ex2)
2484 return ex1->start == ex2->start && ex1->end == ex2->end;
2487 struct ldlm_inodebits {
2490 __u64 try_bits; /* optional bits to try */
2491 __u64 cancel_bits; /* for lock convert */
2495 __u32 li_initiator_id; /* index of MDT that initiated this lock */
2498 struct ldlm_flock_wire {
2506 /* it's important that the fields of the ldlm_extent structure match
2507 * the first fields of the ldlm_flock structure because there is only
2508 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2509 * this ever changes we will need to swab the union differently based
2510 * on the resource type. */
2512 union ldlm_wire_policy_data {
2513 struct ldlm_extent l_extent;
2514 struct ldlm_flock_wire l_flock;
2515 struct ldlm_inodebits l_inodebits;
2518 struct barrier_lvb {
2524 struct ldlm_gl_barrier_desc {
2530 union ldlm_gl_desc {
2531 struct ldlm_gl_lquota_desc lquota_desc;
2532 struct ldlm_gl_barrier_desc barrier_desc;
2535 enum ldlm_intent_flags {
2536 IT_OPEN = 0x00000001,
2537 IT_CREAT = 0x00000002,
2538 IT_OPEN_CREAT = IT_OPEN | IT_CREAT, /* To allow case label. */
2539 IT_READDIR = 0x00000004, /* Used by mdc, not put on the wire. */
2540 IT_GETATTR = 0x00000008,
2541 IT_LOOKUP = 0x00000010,
2542 /* IT_UNLINK = 0x00000020, Obsolete. */
2543 /* IT_TRUNC = 0x00000040, Obsolete. */
2544 IT_GETXATTR = 0x00000080,
2545 /* IT_EXEC = 0x00000100, Obsolete. */
2546 /* IT_PIN = 0x00000200, Obsolete. */
2547 IT_LAYOUT = 0x00000400,
2548 IT_QUOTA_DQACQ = 0x00000800,
2549 IT_QUOTA_CONN = 0x00001000,
2550 /* IT_SETXATTR = 0x00002000, Obsolete. */
2551 IT_GLIMPSE = 0x00004000,
2552 IT_BRW = 0x00008000,
2555 struct ldlm_intent {
2559 struct ldlm_resource_desc {
2560 enum ldlm_type lr_type;
2561 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2562 struct ldlm_res_id lr_name;
2565 struct ldlm_lock_desc {
2566 struct ldlm_resource_desc l_resource;
2567 enum ldlm_mode l_req_mode;
2568 enum ldlm_mode l_granted_mode;
2569 union ldlm_wire_policy_data l_policy_data;
2572 #define LDLM_LOCKREQ_HANDLES 2
2573 #define LDLM_ENQUEUE_CANCEL_OFF 1
2575 struct ldlm_request {
2576 __u32 lock_flags; /* LDLM_FL_*, see lustre_dlm_flags.h */
2577 __u32 lock_count; /* number of locks in lock_handle[] */
2578 struct ldlm_lock_desc lock_desc;/* lock descriptor */
2579 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2584 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2585 struct ldlm_lock_desc lock_desc;
2586 struct lustre_handle lock_handle;
2587 __u64 lock_policy_res1;
2588 __u64 lock_policy_res2;
2591 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2592 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2595 * Opcodes for mountconf (mgs and mgc)
2599 MGS_DISCONNECT = 251,
2600 MGS_EXCEPTION = 252, /* node died, etc. */
2601 MGS_TARGET_REG = 253, /* whenever target starts up */
2602 MGS_TARGET_DEL = 254,
2604 MGS_CONFIG_READ = 256,
2606 MGS_FIRST_OPC = MGS_CONNECT
2609 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 18, 53, 0)
2610 #define MGS_PARAM_MAXLEN 1024
2611 #define KEY_SET_INFO "set_info"
2613 struct mgs_send_param {
2614 char mgs_param[MGS_PARAM_MAXLEN];
2618 /* We pass this info to the MGS so it can write config logs */
2619 #define MTI_NAME_MAXLEN 64
2620 #define MTI_PARAM_MAXLEN 4096
2621 #define MTI_NIDS_MAX 32
2623 struct mgs_target_info {
2624 __u32 mti_lustre_ver;
2625 __u32 mti_stripe_index;
2626 __u32 mti_config_ver;
2627 __u32 mti_flags; /* LDD_F_* */
2628 __u32 mti_nid_count;
2629 __u32 mti_instance; /* Running instance of target */
2630 char mti_fsname[MTI_NAME_MAXLEN];
2631 char mti_svname[MTI_NAME_MAXLEN];
2632 char mti_uuid[sizeof(struct obd_uuid)];
2633 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t) */
2634 char mti_params[MTI_PARAM_MAXLEN];
2635 char mti_nidlist[][LNET_NIDSTR_SIZE];
2636 } __attribute__((packed));
2638 struct mgs_nidtbl_entry {
2639 __u64 mne_version; /* table version of this entry */
2640 __u32 mne_instance; /* target instance # */
2641 __u32 mne_index; /* target index */
2642 __u32 mne_length; /* length of this entry - by bytes */
2643 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2644 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2645 __u8 mne_nid_size; /* size of each NID, by bytes */
2646 __u8 mne_nid_count; /* # of NIDs in buffer */
2648 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2653 MGS_CFG_T_CONFIG = 0,
2654 MGS_CFG_T_SPTLRPC = 1,
2655 MGS_CFG_T_RECOVER = 2,
2656 MGS_CFG_T_PARAMS = 3,
2657 MGS_CFG_T_NODEMAP = 4,
2658 MGS_CFG_T_BARRIER = 5,
2662 struct mgs_config_body {
2663 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2664 __u64 mcb_offset; /* next index of config log to request */
2665 __u16 mcb_type; /* type of log: MGS_CFG_T_[CONFIG|RECOVER] */
2666 __u8 mcb_nm_cur_pass;
2667 __u8 mcb_bits; /* bits unit size of config log */
2668 __u32 mcb_units; /* # of units for bulk transfer */
2671 struct mgs_config_res {
2672 __u64 mcr_offset; /* index of last config log */
2674 __u64 mcr_size; /* size of the log */
2675 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2679 /* Config marker flags (in config log) */
2680 #define CM_START 0x01
2682 #define CM_SKIP 0x04
2683 #define CM_UPGRADE146 0x08
2684 #define CM_EXCLUDE 0x10
2685 #define CM_START_SKIP (CM_START | CM_SKIP)
2688 __u32 cm_step; /* aka config version */
2690 __u32 cm_vers; /* lustre release version number */
2691 __u32 cm_padding; /* 64 bit align */
2692 __s64 cm_createtime; /*when this record was first created */
2693 __s64 cm_canceltime; /*when this record is no longer valid*/
2694 char cm_tgtname[MTI_NAME_MAXLEN];
2695 char cm_comment[MTI_NAME_MAXLEN];
2699 * Opcodes for multiple servers.
2703 /* OBD_LOG_CANCEL = 401, obsolete since 1.5 */
2704 /* OBD_QC_CALLBACK = 402, obsolete since 2.4 */
2707 OBD_FIRST_OPC = OBD_PING
2711 * llog contexts indices.
2713 * There is compatibility problem with indexes below, they are not
2714 * continuous and must keep their numbers for compatibility needs.
2715 * See LU-5218 for details.
2718 LLOG_CONFIG_ORIG_CTXT = 0,
2719 LLOG_CONFIG_REPL_CTXT = 1,
2720 LLOG_MDS_OST_ORIG_CTXT = 2,
2721 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2722 LLOG_SIZE_ORIG_CTXT = 4,
2723 LLOG_SIZE_REPL_CTXT = 5,
2724 LLOG_TEST_ORIG_CTXT = 8,
2725 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2726 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2727 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2728 /* for multiple changelog consumers */
2729 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2730 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2731 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log. reserve for the client */
2732 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log. reserve for the client */
2736 /** Identifier for a single log object */
2738 struct ost_id lgl_oi;
2740 } __attribute__((packed));
2742 /** Records written to the CATALOGS list */
2743 #define CATLIST "CATALOGS"
2745 struct llog_logid lci_logid;
2749 } __attribute__((packed));
2751 /* Log data record types - there is no specific reason that these need to
2752 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2754 #define LLOG_OP_MAGIC 0x10600000
2755 #define LLOG_OP_MASK 0xfff00000
2758 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2759 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2760 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2761 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2762 REINT_UNLINK, /* obsolete after 2.5.0 */
2763 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2765 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2766 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2768 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2769 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2770 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2771 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2772 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2773 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2774 CHANGELOG_USER_REC2 = LLOG_OP_MAGIC | 0x70002,
2775 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2776 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000, /* Resevered to avoid
2779 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2780 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2783 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2784 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2786 /** Log record header - stored in little endian order.
2787 * Each record must start with this struct, end with a llog_rec_tail,
2788 * and be a multiple of 256 bits in size.
2790 struct llog_rec_hdr {
2795 } __attribute__((packed));
2797 struct llog_rec_tail {
2800 } __attribute__((packed));
2802 /* Where data follow just after header */
2803 #define REC_DATA(ptr) \
2804 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2806 #define REC_DATA_LEN(rec) \
2807 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2808 sizeof(struct llog_rec_tail))
2810 struct llog_logid_rec {
2811 struct llog_rec_hdr lid_hdr;
2812 struct llog_logid lid_id;
2816 struct llog_rec_tail lid_tail;
2817 } __attribute__((packed));
2819 struct llog_unlink_rec {
2820 struct llog_rec_hdr lur_hdr;
2824 struct llog_rec_tail lur_tail;
2825 } __attribute__((packed));
2827 struct llog_unlink64_rec {
2828 struct llog_rec_hdr lur_hdr;
2829 struct lu_fid lur_fid;
2830 __u32 lur_count; /* to destroy the lost precreated */
2834 struct llog_rec_tail lur_tail;
2835 } __attribute__((packed));
2837 struct llog_setattr64_rec {
2838 struct llog_rec_hdr lsr_hdr;
2839 struct ost_id lsr_oi;
2845 struct llog_rec_tail lsr_tail;
2846 } __attribute__((packed));
2848 /* Extended to support project quota */
2849 struct llog_setattr64_rec_v2 {
2850 struct llog_rec_hdr lsr_hdr;
2851 struct ost_id lsr_oi;
2858 __u32 lsr_layout_version;
2861 struct llog_rec_tail lsr_tail;
2862 } __attribute__((packed));
2864 struct llog_size_change_rec {
2865 struct llog_rec_hdr lsc_hdr;
2866 struct ll_fid lsc_fid;
2871 struct llog_rec_tail lsc_tail;
2872 } __attribute__((packed));
2874 #define CHANGELOG_MAGIC 0xca103000
2876 /** \a changelog_rec_type's that can't be masked */
2877 #define CHANGELOG_MINMASK BIT(CL_MARK)
2878 /** bits covering all \a changelog_rec_type's */
2879 #define CHANGELOG_ALLMASK (BIT(CL_LAST) - 1)
2880 /** default \a changelog_rec_type mask. Allow all of them, except
2881 * CL_ATIME since it can really be time consuming, and not necessary
2883 * Remove also CL_OPEN, CL_GETXATTR and CL_DN_OPEN from default list as it can
2884 * be costly and only necessary for audit purpose.
2886 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & \
2887 ~(BIT(CL_ATIME) | BIT(CL_OPEN) | BIT(CL_GETXATTR) | BIT(CL_DN_OPEN)))
2889 /* changelog llog name, needed by client replicators */
2890 #define CHANGELOG_CATALOG "changelog_catalog"
2892 struct changelog_setinfo {
2895 } __attribute__((packed));
2897 /** changelog record */
2898 struct llog_changelog_rec {
2899 struct llog_rec_hdr cr_hdr;
2900 struct changelog_rec cr; /**< Variable length field */
2901 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2902 } __attribute__((packed));
2904 #define CHANGELOG_USER_PREFIX "cl"
2905 #define CHANGELOG_USER_NAMELEN 16 /* base name including NUL terminator */
2906 #define CHANGELOG_USER_NAMELEN_FULL 30 /* basename plus 'cl$ID-' prefix */
2908 struct llog_changelog_user_rec {
2909 struct llog_rec_hdr cur_hdr;
2911 /* only intended to be used in relative time comparisons to
2912 * detect idle users */
2915 struct llog_rec_tail cur_tail;
2916 } __attribute__((packed));
2918 /* this is twice the size of CHANGELOG_USER_REC */
2919 struct llog_changelog_user_rec2 {
2920 struct llog_rec_hdr cur_hdr;
2922 /* only for use in relative time comparisons to detect idle users */
2927 char cur_name[CHANGELOG_USER_NAMELEN];
2930 struct llog_rec_tail cur_tail;
2931 } __attribute__((packed));
2933 enum agent_req_status {
2941 static inline const char *agent_req_status2name(enum agent_req_status ars)
2959 struct llog_agent_req_rec {
2960 struct llog_rec_hdr arr_hdr; /**< record header */
2961 __u32 arr_status; /**< status of the request */
2963 * agent_req_status */
2964 __u32 arr_archive_id; /**< backend archive number */
2965 __u64 arr_flags; /**< req flags */
2966 __u64 arr_compound_id; /**< compound cookie, ignored */
2967 __u64 arr_req_create; /**< req. creation time */
2968 __u64 arr_req_change; /**< req. status change time */
2969 struct hsm_action_item arr_hai; /**< req. to the agent */
2970 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2971 } __attribute__((packed));
2973 /* Old llog gen for compatibility */
2977 } __attribute__((packed));
2979 struct llog_gen_rec {
2980 struct llog_rec_hdr lgr_hdr;
2981 struct llog_gen lgr_gen;
2985 struct llog_rec_tail lgr_tail;
2988 /* flags for the logs */
2990 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2991 LLOG_F_IS_CAT = 0x2,
2992 LLOG_F_IS_PLAIN = 0x4,
2993 LLOG_F_EXT_JOBID = 0x8,
2994 LLOG_F_IS_FIXSIZE = 0x10,
2995 LLOG_F_EXT_EXTRA_FLAGS = 0x20,
2996 LLOG_F_EXT_X_UIDGID = 0x40,
2997 LLOG_F_EXT_X_NID = 0x80,
2998 LLOG_F_EXT_X_OMODE = 0x100,
2999 LLOG_F_EXT_X_XATTR = 0x200,
3000 LLOG_F_RM_ON_ERR = 0x400,
3001 LLOG_F_MAX_AGE = 0x800,
3003 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
3004 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
3005 * because the catlog record is usually fixed size, but its plain
3006 * log record can be variable */
3007 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID | LLOG_F_EXT_EXTRA_FLAGS |
3008 LLOG_F_EXT_X_UIDGID | LLOG_F_EXT_X_NID |
3009 LLOG_F_EXT_X_OMODE | LLOG_F_EXT_X_XATTR,
3012 /* On-disk header structure of each log object, stored in little endian order */
3013 #define LLOG_MIN_CHUNK_SIZE 8192
3014 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
3015 * - sizeof(llh_bitmap) */
3016 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
3017 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
3019 struct llog_log_hdr {
3020 struct llog_rec_hdr llh_hdr;
3021 __s64 llh_timestamp;
3023 __u32 llh_bitmap_offset;
3026 /* for a catalog the first/oldest and still in-use plain slot is just
3027 * next to it. It will serve as the upper limit after Catalog has
3030 struct obd_uuid llh_tgtuuid;
3031 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
3032 /* These fields must always be at the end of the llog_log_hdr.
3033 * Note: llh_bitmap size is variable because llog chunk size could be
3034 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
3035 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
3036 * llh_tail should only be refered by LLOG_HDR_TAIL().
3037 * But this structure is also used by client/server llog interface
3038 * (see llog_client.c), it will be kept in its original way to avoid
3039 * compatiblity issue. */
3040 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
3041 struct llog_rec_tail llh_tail;
3042 } __attribute__((packed));
3043 #undef LLOG_HEADER_SIZE
3044 #undef LLOG_BITMAP_BYTES
3046 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
3047 llh->llh_bitmap_offset - \
3048 sizeof(llh->llh_tail)) * 8)
3049 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
3050 (llh)->llh_bitmap_offset)
3051 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
3052 llh->llh_hdr.lrh_len - \
3053 sizeof(llh->llh_tail)))
3055 /** log cookies are used to reference a specific log file and a record therein,
3056 and pass record offset from llog_process_thread to llog_write */
3057 struct llog_cookie {
3059 struct llog_logid lgc_lgl;
3065 } __attribute__((packed));
3067 /** llog protocol */
3068 enum llogd_rpc_ops {
3069 LLOG_ORIGIN_HANDLE_CREATE = 501,
3070 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
3071 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
3072 /* LLOG_ORIGIN_HANDLE_WRITE_REC = 504, Obsolete by 2.1. */
3073 /* LLOG_ORIGIN_HANDLE_CLOSE = 505, Obsolete by 1.8. */
3074 /* LLOG_ORIGIN_CONNECT = 506, Obsolete by 2.4. */
3075 /* LLOG_CATINFO = 507, Obsolete by 2.3. */
3076 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
3077 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* Obsolete by 2.11. */
3079 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
3083 struct llog_logid lgd_logid;
3085 __u32 lgd_llh_flags;
3087 __u32 lgd_saved_index;
3089 __u64 lgd_cur_offset;
3090 } __attribute__((packed));
3092 struct llogd_conn_body {
3093 struct llog_gen lgdc_gen;
3094 struct llog_logid lgdc_logid;
3095 __u32 lgdc_ctxt_idx;
3096 } __attribute__((packed));
3098 /* Note: 64-bit types are 64-bit aligned in structure */
3100 __u64 o_valid; /* hot fields in this obdo */
3103 __u64 o_size; /* o_size-o_blocks == ost_lvb */
3107 __u64 o_blocks; /* brw: cli sent cached bytes */
3110 /* 32-bit fields start here: keep an even number of them via padding */
3111 __u32 o_blksize; /* optimal IO blocksize */
3112 __u32 o_mode; /* brw: cli sent cache remain */
3116 __u32 o_nlink; /* brw: checksum */
3118 __u32 o_misc; /* brw: o_dropped */
3120 __u64 o_ioepoch; /* epoch in ost writes */
3121 __u32 o_stripe_idx; /* holds stripe idx */
3123 struct lustre_handle o_handle; /* brw: lock handle to prolong
3125 /* Originally, the field is llog_cookie for destroy with unlink cookie
3126 * from MDS, it is obsolete in 2.8. Then reuse it by client to transfer
3127 * layout and PFL information in IO, setattr RPCs. Since llog_cookie is
3128 * not used on wire any longer, remove it from the obdo, then it can be
3129 * enlarged freely in the further without affect related RPCs.
3131 * sizeof(ost_layout) + sieof(__u32) == sizeof(llog_cookie). */
3132 struct ost_layout o_layout;
3133 __u32 o_layout_version;
3137 __u64 o_data_version; /* getattr: sum of iversion for
3139 * brw: grant space consumed on
3140 * the client for the write */
3142 __u32 o_padding_4; /* also fix
3143 * lustre_swab_obdo() */
3148 #define o_dirty o_blocks
3149 #define o_undirty o_mode
3150 #define o_dropped o_misc
3151 #define o_cksum o_nlink
3152 #define o_grant_used o_data_version
3153 #define o_falloc_mode o_nlink
3155 struct lfsck_request {
3167 __u16 lr_async_windows;
3169 struct lu_fid lr_fid;
3170 struct lu_fid lr_fid2;
3178 struct lfsck_reply {
3185 LE_LASTID_REBUILDING = 1,
3186 LE_LASTID_REBUILT = 2,
3192 /* LE_FID_ACCESSED = 8, moved to lfsck_events_local */
3194 LE_CONDITIONAL_DESTROY = 10,
3195 LE_PAIRS_VERIFY = 11,
3196 LE_SET_LMV_MASTER = 15,
3197 LE_SET_LMV_SLAVE = 16,
3200 enum lfsck_event_flags {
3201 LEF_TO_OST = 0x00000001,
3202 LEF_FROM_OST = 0x00000002,
3203 LEF_SET_LMV_HASH = 0x00000004,
3204 LEF_SET_LMV_ALL = 0x00000008,
3205 LEF_RECHECK_NAME_HASH = 0x00000010,
3206 LEF_QUERY_ALL = 0x00000020,
3209 /* request structure for OST's */
3214 /* Key for FIEMAP to be used in get_info calls */
3215 struct ll_fiemap_info_key {
3217 struct obdo lfik_oa;
3218 struct fiemap lfik_fiemap;
3221 #define IDX_INFO_MAGIC 0x3D37CC37
3223 /* Index file transfer through the network. The server serializes the index into
3224 * a byte stream which is sent to the client via a bulk transfer */
3228 /* reply: see idx_info_flags below */
3231 /* request & reply: number of lu_idxpage (to be) transferred */
3235 /* request: requested attributes passed down to the iterator API */
3238 /* request & reply: index file identifier (FID) */
3239 struct lu_fid ii_fid;
3241 /* reply: version of the index file before starting to walk the index.
3242 * Please note that the version can be modified at any time during the
3246 /* request: hash to start with:
3247 * reply: hash of the first entry of the first lu_idxpage and hash
3248 * of the entry to read next if any */
3249 __u64 ii_hash_start;
3252 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3256 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3265 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3267 /* List of flags used in idx_info::ii_flags */
3268 enum idx_info_flags {
3269 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3270 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3271 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3272 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3273 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
3276 #define LIP_MAGIC 0x8A6D6B6C
3278 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3280 /* 16-byte header */
3283 __u16 lip_nr; /* number of entries in the container */
3284 __u64 lip_pad0; /* additional padding for future use */
3286 /* key/record pairs are stored in the remaining 4080 bytes.
3287 * depending upon the flags in idx_info::ii_flags, each key/record
3288 * pair might be preceded by:
3290 * - the key size (II_FL_VARKEY is set)
3291 * - the record size (II_FL_VARREC is set)
3293 * For the time being, we only support fixed-size key & record. */
3294 char lip_entries[0];
3297 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3299 /* Gather all possible type associated with a 4KB container */
3301 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3302 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3303 char lp_array[LU_PAGE_SIZE];
3306 /* security opcodes */
3309 SEC_CTX_INIT_CONT = 802,
3312 SEC_FIRST_OPC = SEC_CTX_INIT
3315 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3316 #define LINK_EA_MAGIC 0x11EAF1DFUL
3317 struct link_ea_header {
3320 __u64 leh_len; /* total size */
3321 __u32 leh_overflow_time;
3325 /** Hardlink data is name and parent fid.
3326 * Stored in this crazy struct for maximum packing and endian-neutrality
3328 struct link_ea_entry {
3329 /** __u16 stored big-endian, unaligned */
3330 unsigned char lee_reclen[2];
3331 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3333 } __attribute__((packed));
3335 /** fid2path request/reply structure */
3336 struct getinfo_fid2path {
3337 struct lu_fid gf_fid;
3343 struct lu_fid gf_root_fid[0];
3345 } __attribute__((packed));
3347 /** path2parent request/reply structures */
3349 struct lu_fid gp_fid; /**< parent FID */
3350 __u32 gp_linkno; /**< hardlink number */
3351 __u32 gp_name_size; /**< size of the name field */
3352 char gp_name[0]; /**< zero-terminated link name */
3353 } __attribute__((packed));
3355 enum layout_intent_opc {
3356 LAYOUT_INTENT_ACCESS = 0, /** generic access */
3357 LAYOUT_INTENT_READ = 1, /** not used */
3358 LAYOUT_INTENT_WRITE = 2, /** write file, for comp layout */
3359 LAYOUT_INTENT_GLIMPSE = 3, /** not used */
3360 LAYOUT_INTENT_TRUNC = 4, /** truncate file, for comp layout */
3361 LAYOUT_INTENT_RELEASE = 5, /** reserved for HSM release */
3362 LAYOUT_INTENT_RESTORE = 6, /** reserved for HSM restore */
3363 LAYOUT_INTENT_PCCRO_SET = 7, /** set read-only layout for PCC */
3364 LAYOUT_INTENT_PCCRO_CLEAR = 8, /** clear read-only layout */
3367 /* enqueue layout lock with intent */
3368 struct layout_intent {
3369 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3371 struct lu_extent li_extent;
3372 } __attribute__((packed));
3375 * On the wire version of hsm_progress structure.
3377 * Contains the userspace hsm_progress and some internal fields.
3379 struct hsm_progress_kernel {
3380 /* Field taken from struct hsm_progress */
3381 struct lu_fid hpk_fid;
3383 struct hsm_extent hpk_extent;
3385 __u16 hpk_errval; /* positive val */
3387 /* Additional fields */
3388 __u64 hpk_data_version;
3390 } __attribute__((packed));
3393 * OUT_UPDATE RPC Format
3395 * During the cross-ref operation, the Master MDT, which the client send the
3396 * request to, will disassembly the operation into object updates, then OSP
3397 * will send these updates to the remote MDT to be executed.
3399 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3400 * operation and does a type of modification to an object.
3408 * update (ub_count-th)
3410 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3415 * rc [+ buffers] (1st)
3416 * rc [+ buffers] (2st)
3418 * rc [+ buffers] (nr_count-th)
3420 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3421 * be equal to ub_count.
3425 * Type of each update, if adding/deleting update, please also update
3426 * update_opcode in lustre/target/out_lib.c.
3438 OUT_INDEX_LOOKUP = 9,
3439 OUT_INDEX_INSERT = 10,
3440 OUT_INDEX_DELETE = 11,
3446 OUT_XATTR_LIST = 17,
3451 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3452 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3453 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3454 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3457 struct object_update_param {
3458 __u16 oup_len; /* length of this parameter */
3462 } __attribute__((packed));
3465 struct object_update {
3466 __u16 ou_type; /* enum update_type */
3467 __u16 ou_params_count; /* update parameters count */
3468 __u32 ou_result_size; /* how many bytes can return */
3469 __u32 ou_flags; /* enum update_flag */
3470 __u32 ou_padding1; /* padding 1 */
3471 __u64 ou_batchid; /* op transno on master */
3472 struct lu_fid ou_fid; /* object to be updated */
3473 struct object_update_param ou_params[0]; /* update params */
3476 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3477 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3478 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3479 /* Hold object_updates sending to the remote OUT in single RPC */
3480 struct object_update_request {
3482 __u16 ourq_count; /* number of ourq_updates[] */
3484 struct object_update ourq_updates[0];
3487 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3488 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3489 /* Header for updates request between MDTs */
3490 struct out_update_header {
3493 __u32 ouh_inline_length;
3494 __u32 ouh_reply_size;
3495 __u32 ouh_inline_data[0];
3498 struct out_update_buffer {
3503 /* the result of object update */
3504 struct object_update_result {
3511 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3512 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3513 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3514 /* Hold object_update_results being replied from the remote OUT. */
3515 struct object_update_reply {
3522 /* read update result */
3523 struct out_read_reply {
3530 #define BUT_REQUEST_MAGIC 0xBADE0001
3531 /* Hold batched updates sending to the remote target in a single RPC */
3532 struct batch_update_request {
3533 /* Magic number: BUT_REQUEST_MAGIC. */
3535 /* Number of sub requests packed in this batched RPC: burq_reqmsg[]. */
3537 /* Unused padding field. */
3540 * Sub request message array. As message feild buffers for each sub
3541 * request are packed after padded lustre_msg.lm_buflens[] arrary, thus
3542 * it can locate the next request message via the function
3543 * @batch_update_reqmsg_next() in lustre/include/obj_update.h
3545 struct lustre_msg burq_reqmsg[0];
3548 #define BUT_HEADER_MAGIC 0xBADF0001
3549 /* Header for Batched UpdaTes request */
3550 struct but_update_header {
3551 /* Magic number: BUT_HEADER_MAGIC */
3554 * When the total request buffer length is less than MAX_INLINE_SIZE,
3555 * @buh_count is set with 1 and the batched RPC reqeust can be packed
3557 * Otherwise, @buh_count indicates the IO vector count transferring in
3561 /* inline buffer length when the batched RPC can be packed inline. */
3562 __u32 buh_inline_length;
3563 /* The reply buffer size the client prepared. */
3564 __u32 buh_reply_size;
3565 /* Sub request count in this batched RPC. */
3566 __u32 buh_update_count;
3567 /* Unused padding field. */
3569 /* Inline buffer used when the RPC request can be packed inline. */
3570 __u32 buh_inline_data[0];
3573 struct but_update_buffer {
3578 #define BUT_REPLY_MAGIC 0x00AD0001
3579 /* Batched reply received from a remote targer in a batched RPC. */
3580 struct batch_update_reply {
3581 /* Magic number: BUT_REPLY_MAGIC. */
3583 /* Successful returned sub requests. */
3585 /* Unused padding field. */
3588 * Sub reply message array.
3589 * It can locate the next reply message buffer via the function
3590 * @batch_update_repmsg_next() in lustre/include/obj_update.h
3592 struct lustre_msg burp_repmsg[0];
3596 * Batch update opcode.
3598 enum batch_update_cmd {
3601 BUT_FIRST_OPC = BUT_GETATTR,
3604 /** layout swap request structure
3605 * fid1 and fid2 are in mdt_body
3607 struct mdc_swap_layouts {
3609 } __attribute__((packed));
3611 #define INLINE_RESYNC_ARRAY_SIZE 15
3612 struct close_data_resync_done {
3614 __u32 resync_ids_inline[INLINE_RESYNC_ARRAY_SIZE];
3618 struct lustre_handle cd_handle;
3619 struct lu_fid cd_fid;
3620 __u64 cd_data_version;
3622 __u64 cd_reserved[8];
3623 struct close_data_resync_done cd_resync;
3627 __u32 cd_archive_id;
3631 /* Update llog format */
3633 struct lu_fid uop_fid;
3635 __u16 uop_param_count;
3636 __u16 uop_params_off[];
3637 } __attribute__((packed));
3640 struct update_op uops_op[0];
3643 struct update_params {
3644 struct object_update_param up_params[0];
3647 enum update_records_flag {
3648 UPDATE_RECORD_CONTINUE = 1 >> 0,
3651 * This is the update record format used to store the updates in
3652 * disk. All updates of the operation will be stored in ur_ops.
3653 * All of parameters for updates of the operation will be stored
3655 * To save the space of the record, parameters in ur_ops will only
3656 * remember their offset in ur_params, so to avoid storing duplicate
3657 * parameters in ur_params, which can help us save a lot space for
3658 * operation like creating striped directory.
3660 struct update_records {
3661 __u64 ur_master_transno;
3664 /* If the operation includes multiple updates, then ur_index
3665 * means the index of the update inside the whole updates. */
3667 __u32 ur_update_count;
3668 __u32 ur_param_count;
3669 struct update_ops ur_ops;
3670 /* Note ur_ops has a variable size, so comment out
3671 * the following ur_params, in case some use it directly
3672 * update_records->ur_params
3674 * struct update_params ur_params;
3678 struct llog_update_record {
3679 struct llog_rec_hdr lur_hdr;
3680 struct update_records lur_update_rec;
3681 /* Note ur_update_rec has a variable size, so comment out
3682 * the following ur_tail, in case someone use it directly
3684 * struct llog_rec_tail lur_tail;
3688 /* sepol string format is:
3689 * <1-digit for SELinux status>:<policy name>:<policy version>:<policy hash>
3691 /* Max length of the sepol string
3692 * Should be large enough to contain a sha512sum of the policy
3694 #define SELINUX_MODE_LEN 1
3695 #define SELINUX_POLICY_VER_LEN 3 /* 3 chars to leave room for the future */
3696 #define SELINUX_POLICY_HASH_LEN 64
3697 #define LUSTRE_NODEMAP_SEPOL_LENGTH (SELINUX_MODE_LEN + NAME_MAX + \
3698 SELINUX_POLICY_VER_LEN + \
3699 SELINUX_POLICY_HASH_LEN + 3)
3701 /** enums containing the types of ids contained in a nodemap
3702 * kept so other modules (mgs, mdt, etc) can define the type
3706 enum nodemap_id_type {
3712 enum nodemap_tree_type {
3713 NODEMAP_FS_TO_CLIENT,
3714 NODEMAP_CLIENT_TO_FS,
3717 enum nodemap_mapping_modes {
3718 NODEMAP_MAP_BOTH_LEGACY = 0x0, /* for compatibility */
3719 NODEMAP_MAP_UID = 0x01,
3720 NODEMAP_MAP_GID = 0x02,
3721 NODEMAP_MAP_BOTH = 0x03, /* for compatibility */
3722 NODEMAP_MAP_PROJID = 0x04,
3723 NODEMAP_MAP_ALL = NODEMAP_MAP_UID |
3728 enum nodemap_rbac_roles {
3729 NODEMAP_RBAC_FILE_PERMS = 0x00000001,
3730 NODEMAP_RBAC_DNE_OPS = 0x00000002,
3731 NODEMAP_RBAC_QUOTA_OPS = 0x00000004,
3732 NODEMAP_RBAC_BYFID_OPS = 0x00000008,
3733 NODEMAP_RBAC_CHLG_OPS = 0x00000010,
3734 NODEMAP_RBAC_FSCRYPT_ADMIN = 0x00000020,
3735 NODEMAP_RBAC_NONE = (__u32)~(NODEMAP_RBAC_FILE_PERMS |
3736 NODEMAP_RBAC_DNE_OPS |
3737 NODEMAP_RBAC_QUOTA_OPS |
3738 NODEMAP_RBAC_BYFID_OPS |
3739 NODEMAP_RBAC_CHLG_OPS |
3740 NODEMAP_RBAC_FSCRYPT_ADMIN),
3741 NODEMAP_RBAC_ALL = 0xFFFFFFFF, /* future caps ON by default */
3745 * rawobj stuff for GSS
3747 typedef struct netobj_s {
3752 typedef struct rawobj_s {
3759 * following 3 headers must have the same sizes and offsets
3762 __u8 gh_version; /* gss version */
3763 __u8 gh_sp; /* sec part */
3765 __u32 gh_flags; /* wrap flags */
3766 __u32 gh_proc; /* proc */
3767 __u32 gh_seq; /* sequence */
3768 __u32 gh_svc; /* service */
3772 netobj_t gh_handle; /* context handle */
3775 struct gss_rep_header {
3789 struct gss_err_header {
3804 * GSS part of wire context information sent from client, saved and
3805 * used later by server.
3807 struct gss_wire_ctx {
3815 /* This is the lu_ladvise struct which goes out on the wire.
3816 * Corresponds to the userspace arg llapi_lu_ladvise.
3817 * value[1-4] are unspecified fields, used differently by different advices */
3819 __u16 lla_advice; /* advice type */
3820 __u16 lla_value1; /* values for different advice types */
3822 __u64 lla_start; /* first byte of extent for advice */
3823 __u64 lla_end; /* last byte of extent for advice */
3828 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3829 * arg llapi_ladvise_hdr.
3830 * value[1-3] are unspecified fields, used differently by different advices */
3831 struct ladvise_hdr {
3832 __u32 lah_magic; /* LADVISE_MAGIC */
3833 __u32 lah_count; /* number of advices */
3834 __u64 lah_flags; /* from enum ladvise_flag */
3835 __u32 lah_value1; /* unused */
3836 __u32 lah_value2; /* unused */
3837 __u64 lah_value3; /* unused */
3838 struct lu_ladvise lah_advise[0]; /* advices in this header */
3841 #if defined(__cplusplus)