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 __u64 pb_padding64_2;
726 /* message body offset for lustre_msg_v2 */
727 /* ptlrpc body offset in all request/reply messages */
728 #define MSG_PTLRPC_BODY_OFF 0
730 /* normal request/reply message record offset */
731 #define REQ_REC_OFF 1
732 #define REPLY_REC_OFF 1
734 /* ldlm request message body offset */
735 #define DLM_LOCKREQ_OFF 1 /* lockreq offset */
736 #define DLM_REQ_REC_OFF 2 /* normal dlm request record offset */
738 /* ldlm intent lock message body offset */
739 #define DLM_INTENT_IT_OFF 2 /* intent lock it offset */
740 #define DLM_INTENT_REC_OFF 3 /* intent lock record offset */
742 /* ldlm reply message body offset */
743 #define DLM_LOCKREPLY_OFF 1 /* lockrep offset */
744 #define DLM_REPLY_REC_OFF 2 /* reply record offset */
746 /** only use in req->rq_{req,rep}_swab_mask */
747 #define MSG_PTLRPC_HEADER_OFF 31
750 #define OBD_CONNECT_RDONLY 0x1ULL /* client is read-only */
751 #define OBD_CONNECT_INDEX 0x2ULL /* connect to LOV idx */
752 #define OBD_CONNECT_MDS 0x4ULL /* connect MDT to OST */
753 #define OBD_CONNECT_GRANT 0x8ULL /* fetch grant connect */
754 #define OBD_CONNECT_SRVLOCK 0x10ULL /* server lock for RPC */
755 #define OBD_CONNECT_VERSION 0x20ULL /* versions in OCD */
756 #define OBD_CONNECT_REQPORTAL 0x40ULL /* non-IO portal */
757 #define OBD_CONNECT_ACL 0x80ULL /* access control list */
758 #define OBD_CONNECT_XATTR 0x100ULL /* extended attributes */
759 #define OBD_CONNECT_LARGE_ACL 0x200ULL /* over 32 ACL entries */
760 /* was OBD_CONNECT_TRUNCLOCK 0x400ULL * server locks punch */
761 /* temporary reuse until 2.21.53 to indicate pre-2.15 client, see LU-15478 */
762 #define OBD_CONNECT_OLD_FALLOC 0x400ULL /* no o_valid flags */
763 #define OBD_CONNECT_TRANSNO 0x800ULL /* replay send transno */
764 #define OBD_CONNECT_IBITS 0x1000ULL /* not checked 2.11+ */
765 #define OBD_CONNECT_BARRIER 0x2000ULL /* write barrier */
766 #define OBD_CONNECT_ATTRFID 0x4000ULL /* Server GetAttr FID */
767 #define OBD_CONNECT_NODEVOH 0x8000ULL /* No open handle spec */
768 #define OBD_CONNECT_RMT_CLIENT 0x10000ULL /* Never used, gone 2.9*/
769 #define OBD_CONNECT_RMT_CLIENT_FORCE 0x20000ULL /* Never used, gone 2.9*/
770 #define OBD_CONNECT_BRW_SIZE 0x40000ULL /* Max bytes per rpc */
771 #define OBD_CONNECT_QUOTA64 0x80000ULL /* Unused since 2.4 */
772 #define OBD_CONNECT_MDS_CAPA 0x100000ULL /* Unused since 2.7 */
773 #define OBD_CONNECT_OSS_CAPA 0x200000ULL /* Unused since 2.7 */
774 #define OBD_CONNECT_CANCELSET 0x400000ULL /* Early batch cancel */
775 #define OBD_CONNECT_SOM 0x800000ULL /* Unused since 2.7 */
776 #define OBD_CONNECT_AT 0x1000000ULL /* client uses AT */
777 #define OBD_CONNECT_LRU_RESIZE 0x2000000ULL /* LRU resize feature */
778 #define OBD_CONNECT_MDS_MDS 0x4000000ULL /* MDS-MDS connection */
779 #define OBD_CONNECT_REAL 0x8000000ULL /* Unused since 2.8 */
780 #define OBD_CONNECT_CHANGE_QS 0x10000000ULL /* Unused since 2.4 */
781 #define OBD_CONNECT_CKSUM 0x20000000ULL /* cksum algo choice */
782 #define OBD_CONNECT_FID 0x40000000ULL /* server handles FIDs */
783 #define OBD_CONNECT_VBR 0x80000000ULL /* version based recov */
784 #define OBD_CONNECT_LOV_V3 0x100000000ULL /* client LOV v3 EA */
785 #define OBD_CONNECT_GRANT_SHRINK 0x200000000ULL /* handle grant shrink */
786 #define OBD_CONNECT_SKIP_ORPHAN 0x400000000ULL /* no orph objid reuse */
787 #define OBD_CONNECT_MAX_EASIZE 0x800000000ULL /* EA size in reply */
788 #define OBD_CONNECT_FULL20 0x1000000000ULL /* it is 2.0 client */
789 #define OBD_CONNECT_LAYOUTLOCK 0x2000000000ULL /* client layout lock */
790 #define OBD_CONNECT_64BITHASH 0x4000000000ULL /* 64-bits dir hash */
791 #define OBD_CONNECT_MAXBYTES 0x8000000000ULL /* max stripe size */
792 #define OBD_CONNECT_IMP_RECOV 0x10000000000ULL /* imp recov support */
793 #define OBD_CONNECT_JOBSTATS 0x20000000000ULL /* ptlrpc_body jobid */
794 #define OBD_CONNECT_UMASK 0x40000000000ULL /* create client umask */
795 #define OBD_CONNECT_EINPROGRESS 0x80000000000ULL /* client -EINPROGRESS
796 * RPC error handling */
797 #define OBD_CONNECT_GRANT_PARAM 0x100000000000ULL /* extra grant params for
798 * space reservation */
799 #define OBD_CONNECT_FLOCK_OWNER 0x200000000000ULL /* unused since 2.0 */
800 #define OBD_CONNECT_LVB_TYPE 0x400000000000ULL /* variable LVB type */
801 #define OBD_CONNECT_NANOSEC_TIME 0x800000000000ULL /* nanosec timestamp */
802 #define OBD_CONNECT_LIGHTWEIGHT 0x1000000000000ULL /* lightweight connect */
803 #define OBD_CONNECT_SHORTIO 0x2000000000000ULL /* short io */
804 #define OBD_CONNECT_PINGLESS 0x4000000000000ULL /* pings not required */
805 #define OBD_CONNECT_FLOCK_DEAD 0x8000000000000ULL /* flk deadlock detect */
806 #define OBD_CONNECT_DISP_STRIPE 0x10000000000000ULL /* create stripe disp */
807 #define OBD_CONNECT_OPEN_BY_FID 0x20000000000000ULL /* open by FID won't pack
809 #define OBD_CONNECT_LFSCK 0x40000000000000ULL /* allow online LFSCK */
810 #define OBD_CONNECT_UNLINK_CLOSE 0x100000000000000ULL /* unlink closes file */
811 #define OBD_CONNECT_MULTIMODRPCS 0x200000000000000ULL /* allow multiple change
812 * RPCs in parallel */
813 #define OBD_CONNECT_DIR_STRIPE 0x400000000000000ULL /* striped DNE dir */
814 #define OBD_CONNECT_SUBTREE 0x800000000000000ULL /* fileset mount */
815 /* was OBD_CONNECT_LOCKAHEAD_OLD 0x1000000000000000ULL old lockahead 2.12-2.13*/
816 #define OBD_CONNECT_BULK_MBITS 0x2000000000000000ULL /* ptlrpc_body matchbit*/
817 #define OBD_CONNECT_OBDOPACK 0x4000000000000000ULL /* compact OUT obdo */
818 #define OBD_CONNECT_FLAGS2 0x8000000000000000ULL /* second flags word */
819 /* ocd_connect_flags2 flags */
820 #define OBD_CONNECT2_FILE_SECCTX 0x1ULL /* security context */
821 #define OBD_CONNECT2_LOCKAHEAD 0x2ULL /* ladvise lockahead */
822 #define OBD_CONNECT2_DIR_MIGRATE 0x4ULL /* migrate striped dir */
823 #define OBD_CONNECT2_SUM_STATFS 0x8ULL /* MDT aggregate statfs*/
824 #define OBD_CONNECT2_OVERSTRIPING 0x10ULL /* OST overstriping */
825 #define OBD_CONNECT2_FLR 0x20ULL /* FLR mirror handling */
826 #define OBD_CONNECT2_WBC_INTENTS 0x40ULL /* MDS wb cache intent */
827 #define OBD_CONNECT2_LOCK_CONVERT 0x80ULL /* IBITS lock convert */
828 #define OBD_CONNECT2_ARCHIVE_ID_ARRAY 0x100ULL /* HSM archive_id array*/
829 #define OBD_CONNECT2_INC_XID 0x200ULL /* Increasing xid */
830 #define OBD_CONNECT2_SELINUX_POLICY 0x400ULL /* cli SELinux policy */
831 #define OBD_CONNECT2_LSOM 0x800ULL /* Lazy Size on MDT */
832 #define OBD_CONNECT2_PCC 0x1000ULL /* Persist Client Cache*/
833 #define OBD_CONNECT2_CRUSH 0x2000ULL /* CRUSH dir hash */
834 #define OBD_CONNECT2_ASYNC_DISCARD 0x4000ULL /* async DoM discard */
835 #define OBD_CONNECT2_ENCRYPT 0x8000ULL /* client disk encrypt */
836 #define OBD_CONNECT2_FIDMAP 0x10000ULL /* MDT migrate FID map */
837 #define OBD_CONNECT2_GETATTR_PFID 0x20000ULL /* parent FID getattr */
838 #define OBD_CONNECT2_LSEEK 0x40000ULL /* SEEK_HOLE/DATA RPC */
839 #define OBD_CONNECT2_DOM_LVB 0x80000ULL /* DoM glimpse in LVB */
840 #define OBD_CONNECT2_REP_MBITS 0x100000ULL /* reply mbits, not XID*/
841 #define OBD_CONNECT2_MODE_CONVERT 0x200000ULL /* LDLM mode convert */
842 #define OBD_CONNECT2_BATCH_RPC 0x400000ULL /* Multi-op batch RPCs */
843 #define OBD_CONNECT2_PCCRO 0x800000ULL /* PCC read-only */
844 #define OBD_CONNECT2_MNE_TYPE 0x1000000ULL /* mne_nid_type IPv6 */
845 #define OBD_CONNECT2_LOCK_CONTENTION 0x2000000ULL /* contention detect */
846 #define OBD_CONNECT2_ATOMIC_OPEN_LOCK 0x4000000ULL /* lock on first open */
847 #define OBD_CONNECT2_ENCRYPT_NAME 0x8000000ULL /* name encrypt */
848 #define OBD_CONNECT2_DMV_IMP_INHERIT 0x20000000ULL /* client handle DMV inheritance */
849 #define OBD_CONNECT2_ENCRYPT_FID2PATH 0x40000000ULL /* fid2path enc file */
850 /* For MDS+OSS rolling upgrade interop with 2.16+older, ignored after 2.20.53 */
851 #define OBD_CONNECT2_REPLAY_CREATE 0x80000000ULL /* replay OST_CREATE */
852 #define OBD_CONNECT2_LARGE_NID 0x100000000ULL /* understands large/IPv6 NIDs */
853 #define OBD_CONNECT2_COMPRESS 0x200000000ULL /* compressed file */
854 /* XXX README XXX README XXX README XXX README XXX README XXX README XXX
855 * Please DO NOT add OBD_CONNECT flags before first ensuring that this value
856 * is not in use by some other branch/patch. Email adilger@whamcloud.com
857 * to reserve the new OBD_CONNECT value for use by your feature. Then, submit
858 * a small patch against master and LTS branches that ONLY adds the new flag,
859 * updates obd_connect_names[], adds the flag to check_obd_connect_data(),
860 * and updates wiretests accordingly, so it can be approved and landed easily
861 * to reserve the flag for future use by your feature (submitted separately).
864 #define OCD_HAS_FLAG(ocd, flag) \
865 (!!((ocd)->ocd_connect_flags & OBD_CONNECT_##flag))
866 #define OCD_HAS_FLAG2(ocd, flag2) (OCD_HAS_FLAG(ocd, FLAGS2) && \
867 !!((ocd)->ocd_connect_flags2 & OBD_CONNECT2_##flag2))
870 #ifdef HAVE_LRU_RESIZE_SUPPORT
871 #define LRU_RESIZE_CONNECT_FLAG OBD_CONNECT_LRU_RESIZE
873 #define LRU_RESIZE_CONNECT_FLAG 0
876 #define MDT_CONNECT_SUPPORTED (OBD_CONNECT_RDONLY | OBD_CONNECT_VERSION | \
877 OBD_CONNECT_ACL | OBD_CONNECT_XATTR | \
878 OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH | \
879 OBD_CONNECT_ATTRFID | OBD_CONNECT_CANCELSET | \
880 OBD_CONNECT_AT | OBD_CONNECT_BRW_SIZE | \
881 OBD_CONNECT_MDS_MDS | OBD_CONNECT_FID | \
882 LRU_RESIZE_CONNECT_FLAG | OBD_CONNECT_VBR | \
883 OBD_CONNECT_LOV_V3 | OBD_CONNECT_FULL20 | \
884 OBD_CONNECT_64BITHASH | OBD_CONNECT_JOBSTATS | \
885 OBD_CONNECT_EINPROGRESS | \
886 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_UMASK | \
887 OBD_CONNECT_LVB_TYPE | OBD_CONNECT_LAYOUTLOCK |\
888 OBD_CONNECT_PINGLESS | OBD_CONNECT_MAX_EASIZE |\
889 OBD_CONNECT_FLOCK_DEAD | \
890 OBD_CONNECT_DISP_STRIPE | OBD_CONNECT_LFSCK | \
891 OBD_CONNECT_OPEN_BY_FID | \
892 OBD_CONNECT_DIR_STRIPE | OBD_CONNECT_GRANT | \
893 OBD_CONNECT_SRVLOCK | OBD_CONNECT_BULK_MBITS |\
895 OBD_CONNECT_MULTIMODRPCS |\
896 OBD_CONNECT_SUBTREE | OBD_CONNECT_LARGE_ACL |\
897 OBD_CONNECT_GRANT_PARAM | \
898 OBD_CONNECT_GRANT_SHRINK | \
899 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
901 #define MDT_CONNECT_SUPPORTED2 (OBD_CONNECT2_FILE_SECCTX | \
902 OBD_CONNECT2_DIR_MIGRATE | \
903 OBD_CONNECT2_SUM_STATFS | \
904 OBD_CONNECT2_OVERSTRIPING | \
906 OBD_CONNECT2_LOCK_CONVERT | \
907 OBD_CONNECT2_ARCHIVE_ID_ARRAY | \
908 OBD_CONNECT2_INC_XID | \
909 OBD_CONNECT2_SELINUX_POLICY | \
910 OBD_CONNECT2_LSOM | \
911 OBD_CONNECT2_ASYNC_DISCARD | \
913 OBD_CONNECT2_CRUSH | \
914 OBD_CONNECT2_ENCRYPT | \
915 OBD_CONNECT2_GETATTR_PFID |\
916 OBD_CONNECT2_LSEEK | OBD_CONNECT2_DOM_LVB |\
917 OBD_CONNECT2_REP_MBITS | \
918 OBD_CONNECT2_ATOMIC_OPEN_LOCK | \
919 OBD_CONNECT2_BATCH_RPC | \
920 OBD_CONNECT2_ENCRYPT_NAME | \
921 OBD_CONNECT2_ENCRYPT_FID2PATH | \
922 OBD_CONNECT2_DMV_IMP_INHERIT)
924 #define OST_CONNECT_SUPPORTED (OBD_CONNECT_SRVLOCK | OBD_CONNECT_GRANT | \
925 OBD_CONNECT_REQPORTAL | OBD_CONNECT_VERSION | \
926 OBD_CONNECT_INDEX | \
927 OBD_CONNECT_BRW_SIZE | OBD_CONNECT_CANCELSET | \
928 OBD_CONNECT_AT | LRU_RESIZE_CONNECT_FLAG | \
929 OBD_CONNECT_CKSUM | OBD_CONNECT_VBR | \
930 OBD_CONNECT_MDS | OBD_CONNECT_SKIP_ORPHAN | \
931 OBD_CONNECT_GRANT_SHRINK | OBD_CONNECT_FULL20 |\
932 OBD_CONNECT_64BITHASH | OBD_CONNECT_MAXBYTES | \
933 OBD_CONNECT_MAX_EASIZE | \
934 OBD_CONNECT_EINPROGRESS | \
935 OBD_CONNECT_JOBSTATS | \
936 OBD_CONNECT_LIGHTWEIGHT | OBD_CONNECT_LVB_TYPE|\
937 OBD_CONNECT_LAYOUTLOCK | OBD_CONNECT_FID | \
938 OBD_CONNECT_PINGLESS | OBD_CONNECT_LFSCK | \
939 OBD_CONNECT_BULK_MBITS | \
940 OBD_CONNECT_GRANT_PARAM | \
941 OBD_CONNECT_SHORTIO | OBD_CONNECT_FLAGS2)
943 #define OST_CONNECT_SUPPORTED2 (OBD_CONNECT2_LOCKAHEAD | OBD_CONNECT2_INC_XID |\
944 OBD_CONNECT2_ENCRYPT | OBD_CONNECT2_LSEEK |\
945 OBD_CONNECT2_REP_MBITS |\
946 OBD_CONNECT2_REPLAY_CREATE)
948 #define ECHO_CONNECT_SUPPORTED (OBD_CONNECT_FID | OBD_CONNECT_FLAGS2)
949 #define ECHO_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS
951 #define MGS_CONNECT_SUPPORTED (OBD_CONNECT_VERSION | OBD_CONNECT_AT | \
952 OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV | \
953 OBD_CONNECT_PINGLESS |\
954 OBD_CONNECT_BULK_MBITS | OBD_CONNECT_BARRIER | \
957 #define MGS_CONNECT_SUPPORTED2 OBD_CONNECT2_REP_MBITS | \
958 OBD_CONNECT2_LARGE_NID
960 /* Features required for this version of the client to work with server */
961 #define CLIENT_CONNECT_MDT_REQD (OBD_CONNECT_FID | \
962 OBD_CONNECT_ATTRFID | \
965 /* This structure is used for both request and reply.
967 * If we eventually have separate connect data for different types, which we
968 * almost certainly will, then perhaps we stick a union in here. */
969 struct obd_connect_data {
970 __u64 ocd_connect_flags; /* OBD_CONNECT_* per above */
971 __u32 ocd_version; /* lustre release version number */
972 __u32 ocd_grant; /* initial cache grant amount (bytes) */
973 __u32 ocd_index; /* LOV index to connect to */
974 __u32 ocd_brw_size; /* Maximum BRW size in bytes */
975 __u64 ocd_ibits_known; /* inode bits this client understands */
976 __u8 ocd_grant_blkbits; /* log2 of the backend filesystem blocksize */
977 __u8 ocd_grant_inobits; /* log2 of the per-inode space consumption */
978 __u16 ocd_grant_tax_kb; /* extent insertion overhead, in 1K blocks */
979 __u32 ocd_grant_max_blks;/* maximum number of blocks per extent */
980 __u64 ocd_transno; /* first transno from client to be replayed */
981 __u32 ocd_group; /* MDS group on OST */
982 __u32 ocd_cksum_types; /* supported checksum algorithms */
983 __u32 ocd_max_easize; /* How big LOV EA can be on MDS */
984 __u32 ocd_instance; /* instance # of this target */
985 __u64 ocd_maxbytes; /* Maximum stripe size in bytes */
986 /* Fields after ocd_maxbytes are only accessible by the receiver
987 * if the corresponding flag in ocd_connect_flags is set. Accessing
988 * any field after ocd_maxbytes on the receiver without a valid flag
989 * may result in out-of-bound memory access and kernel oops. */
990 __u16 ocd_maxmodrpcs; /* Maximum modify RPCs in parallel */
991 __u16 padding0; /* READ BELOW! also fix lustre_swab_connect */
992 __u32 padding1; /* READ BELOW! also fix lustre_swab_connect */
993 __u64 ocd_connect_flags2;/* OBD_CONNECT2_* per above */
994 __u64 ocd_compr_type; /* bitmask of supported compression types */
995 __u64 padding4; /* READ BELOW! also fix lustre_swab_connect */
996 __u64 padding5; /* READ BELOW! also fix lustre_swab_connect */
997 __u64 padding6; /* READ BELOW! also fix lustre_swab_connect */
998 __u64 padding7; /* READ BELOW! also fix lustre_swab_connect */
999 __u64 padding8; /* READ BELOW! also fix lustre_swab_connect */
1000 __u64 padding9; /* READ BELOW! also fix lustre_swab_connect */
1001 __u64 paddingA; /* READ BELOW! also fix lustre_swab_connect */
1002 __u64 paddingB; /* READ BELOW! also fix lustre_swab_connect */
1003 __u64 paddingC; /* READ BELOW! also fix lustre_swab_connect */
1004 __u64 paddingD; /* READ BELOW! also fix lustre_swab_connect */
1005 __u64 paddingE; /* READ BELOW! also fix lustre_swab_connect */
1006 __u64 paddingF; /* added 2.1.0. also fix lustre_swab_connect */
1008 /* XXX README XXX README XXX README XXX README XXX README XXX README XXX
1009 * Please DO NOT use any fields before first ensuring that this field is
1010 * not in use by some other branch/patch. Email adilger@whamcloud.com to
1011 * reserve the new obd_connect_data field for use by your feature. Then, submit
1012 * a small patch against master and LTS branch that ONLY adds the new field,
1013 * updates lustre_swab_connect(), along with the matching OBD_CONNECT flag,
1014 * and updates wiretests accordingly,so it can be approved and landed easily
1015 * to reserve the field for future use by your feature (submitted separately).
1019 * Supported checksum algorithms. Up to 32 checksum types are supported.
1020 * (32-bit mask stored in obd_connect_data::ocd_cksum_types)
1021 * Please update DECLARE_CKSUM_NAME in obd_cksum.h when adding a new
1022 * algorithm and also the OBD_FL_CKSUM* flags, OBD_CKSUM_ALL flag,
1023 * OBD_FL_CKSUM_ALL flag and potentially OBD_CKSUM_T10_ALL flag.
1026 OBD_CKSUM_CRC32 = 0x00000001,
1027 OBD_CKSUM_ADLER = 0x00000002,
1028 OBD_CKSUM_CRC32C = 0x00000004,
1029 OBD_CKSUM_RESERVED = 0x00000008,
1030 OBD_CKSUM_T10IP512 = 0x00000010,
1031 OBD_CKSUM_T10IP4K = 0x00000020,
1032 OBD_CKSUM_T10CRC512 = 0x00000040,
1033 OBD_CKSUM_T10CRC4K = 0x00000080,
1036 #define OBD_CKSUM_T10_ALL (OBD_CKSUM_T10IP512 | OBD_CKSUM_T10IP4K | \
1037 OBD_CKSUM_T10CRC512 | OBD_CKSUM_T10CRC4K)
1039 #define OBD_CKSUM_ALL (OBD_CKSUM_CRC32 | OBD_CKSUM_ADLER | OBD_CKSUM_CRC32C | \
1043 * The default checksum algorithm used on top of T10PI GRD tags for RPC.
1044 * Considering that the checksum-of-checksums is only computing CRC32 on a
1045 * 4KB chunk of GRD tags for a 1MB RPC for 512B sectors, or 16KB of GRD
1046 * tags for 16MB of 4KB sectors, this is only 1/256 or 1/1024 of the
1047 * total data being checksummed, so the checksum type used here should not
1048 * affect overall system performance noticeably.
1050 #define OBD_CKSUM_T10_TOP OBD_CKSUM_ADLER
1053 * OST requests: OBDO & OBD request records
1058 OST_REPLY = 0, /* reply ? */
1074 OST_QUOTACHECK = 18, /* not used since 2.4 */
1076 OST_QUOTA_ADJUST_QUNIT = 20, /* not used since 2.4 */
1080 OST_LAST_OPC /* must be < 33 to avoid MDS_GETATTR */
1082 #define OST_FIRST_OPC OST_REPLY
1085 OBD_FL_INLINEDATA = 0x00000001,
1086 OBD_FL_OBDMDEXISTS = 0x00000002,
1087 OBD_FL_DELORPHAN = 0x00000004, /* if set in o_flags delete orphans */
1088 OBD_FL_NORPC = 0x00000008, /* set in o_flags do in OSC not OST */
1089 OBD_FL_IDONLY = 0x00000010, /* set in o_flags only adjust obj id*/
1090 OBD_FL_RECREATE_OBJS= 0x00000020, /* recreate missing obj */
1091 OBD_FL_DEBUG_CHECK = 0x00000040, /* echo client/server debug check */
1092 OBD_FL_NO_PRJQUOTA = 0x00000080, /* the object's project is over
1094 OBD_FL_NO_USRQUOTA = 0x00000100, /* the object's owner is over quota */
1095 OBD_FL_NO_GRPQUOTA = 0x00000200, /* the object's group is over quota */
1096 OBD_FL_CREATE_CROW = 0x00000400, /* object should be create on write */
1097 OBD_FL_SRVLOCK = 0x00000800, /* delegate DLM locking to server */
1098 OBD_FL_CKSUM_CRC32 = 0x00001000, /* CRC32 checksum type */
1099 OBD_FL_CKSUM_ADLER = 0x00002000, /* ADLER checksum type */
1100 OBD_FL_CKSUM_CRC32C = 0x00004000, /* CRC32C checksum type */
1101 OBD_FL_CKSUM_T10IP512 = 0x00005000, /* T10PI IP cksum, 512B sector */
1102 OBD_FL_CKSUM_T10IP4K = 0x00006000, /* T10PI IP cksum, 4KB sector */
1103 OBD_FL_CKSUM_T10CRC512 = 0x00007000, /* T10PI CRC cksum, 512B sector */
1104 OBD_FL_CKSUM_T10CRC4K = 0x00008000, /* T10PI CRC cksum, 4KB sector */
1105 OBD_FL_CKSUM_RSVD3 = 0x00010000, /* for future cksum types */
1106 OBD_FL_SHRINK_GRANT = 0x00020000, /* object shrink the grant */
1107 OBD_FL_MMAP = 0x00040000, /* object is mmapped on the client.
1108 * XXX: obsoleted - reserved for old
1109 * clients prior than 2.2 */
1110 OBD_FL_RECOV_RESEND = 0x00080000, /* recoverable resent */
1111 OBD_FL_NOSPC_BLK = 0x00100000, /* no more block space on OST */
1112 OBD_FL_FLUSH = 0x00200000, /* flush pages on the OST */
1113 OBD_FL_SHORT_IO = 0x00400000, /* short io request */
1114 OBD_FL_ROOT_SQUASH = 0x00800000, /* root squash */
1115 OBD_FL_ROOT_PRJQUOTA = 0x01000000, /* check prj quota for root */
1116 /* OBD_FL_LOCAL_MASK = 0xF0000000, was local-only flags until 2.10 */
1119 * Note that while the original checksum values were separate bits,
1120 * in 2.x we can actually allow all values from 1-31. T10-PI checksum
1121 * types already use values which are not separate bits.
1123 OBD_FL_CKSUM_ALL = OBD_FL_CKSUM_CRC32 | OBD_FL_CKSUM_ADLER |
1124 OBD_FL_CKSUM_CRC32C | OBD_FL_CKSUM_T10IP512 |
1125 OBD_FL_CKSUM_T10IP4K | OBD_FL_CKSUM_T10CRC512 |
1126 OBD_FL_CKSUM_T10CRC4K,
1128 OBD_FL_NO_QUOTA_ALL = OBD_FL_NO_USRQUOTA | OBD_FL_NO_GRPQUOTA |
1133 * All LOV EA magics should have the same postfix, if some new version
1134 * Lustre instroduces new LOV EA magic, then when down-grade to an old
1135 * Lustre, even though the old version system does not recognizes such
1136 * new magic, it still can distinguish the corrupted cases by checking
1137 * the magic's postfix.
1139 #define LOV_MAGIC_MAGIC 0x0BD0
1140 #define LOV_MAGIC_MASK 0xFFFF
1142 #define LOV_MAGIC_V1 (0x0BD10000 | LOV_MAGIC_MAGIC)
1143 #define LOV_MAGIC_JOIN_V1 (0x0BD20000 | LOV_MAGIC_MAGIC)
1144 #define LOV_MAGIC_V3 (0x0BD30000 | LOV_MAGIC_MAGIC)
1145 #define LOV_MAGIC_MIGRATE (0x0BD40000 | LOV_MAGIC_MAGIC)
1146 /* reserved for specifying OSTs */
1147 #define LOV_MAGIC_SPECIFIC (0x0BD50000 | LOV_MAGIC_MAGIC)
1148 #define LOV_MAGIC LOV_MAGIC_V1
1149 #define LOV_MAGIC_COMP_V1 (0x0BD60000 | LOV_MAGIC_MAGIC)
1150 #define LOV_MAGIC_FOREIGN (0x0BD70000 | LOV_MAGIC_MAGIC)
1151 #define LOV_MAGIC_SEL (0x0BD80000 | LOV_MAGIC_MAGIC)
1154 * magic for fully defined striping
1155 * the idea is that we should have different magics for striping "hints"
1156 * (struct lov_user_md_v[13]) and defined ready-to-use striping (struct
1157 * lov_mds_md_v[13]). at the moment the magics are used in wire protocol,
1158 * we can't just change it w/o long way preparation, but we still need a
1159 * mechanism to allow LOD to differentiate hint versus ready striping.
1160 * so, at the moment we do a trick: MDT knows what to expect from request
1161 * depending on the case (replay uses ready striping, non-replay req uses
1162 * hints), so MDT replaces magic with appropriate one and now LOD can
1163 * easily understand what's inside -bzzz
1165 * those *_DEF magics are only used on server side internally, they
1166 * won't be put on wire or disk.
1168 #define LOV_MAGIC_DEFINED 0x10000000
1169 #define LOV_MAGIC_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V1)
1170 #define LOV_MAGIC_V3_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_V3)
1171 #define LOV_MAGIC_COMP_V1_DEFINED (LOV_MAGIC_DEFINED | LOV_MAGIC_COMP_V1)
1173 #define lov_pattern(pattern) (pattern & ~LOV_PATTERN_F_MASK)
1174 #define lov_pattern_flags(pattern) (pattern & LOV_PATTERN_F_MASK)
1176 #define lov_ost_data lov_ost_data_v1
1177 struct lov_ost_data_v1 { /* per-stripe data structure (little-endian)*/
1178 struct ost_id l_ost_oi; /* OST object ID */
1179 __u32 l_ost_gen; /* generation of this l_ost_idx */
1180 __u32 l_ost_idx; /* OST index in LOV (lov_tgt_desc->tgts) */
1183 #define lov_mds_md lov_mds_md_v1
1184 struct lov_mds_md_v1 { /* LOV EA mds/wire data (little-endian) */
1185 __u32 lmm_magic; /* magic number = LOV_MAGIC_V1 */
1186 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1187 struct ost_id lmm_oi; /* LOV object ID */
1188 __u32 lmm_stripe_size; /* size of stripe in bytes */
1189 /* lmm_stripe_count used to be __u32 */
1190 __u16 lmm_stripe_count; /* num stripes in use for this object */
1191 __u16 lmm_layout_gen; /* layout generation number */
1192 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1195 #define MAX_MD_SIZE_OLD (sizeof(struct lov_mds_md) + \
1196 4 * sizeof(struct lov_ost_data))
1197 #define MAX_MD_SIZE (sizeof(struct lov_comp_md_v1) + \
1198 4 * (sizeof(struct lov_comp_md_entry_v1) + \
1200 #define MIN_MD_SIZE (sizeof(struct lov_mds_md) + 1 * sizeof(struct lov_ost_data))
1202 /* This is the default MDT reply size allocated, should the striping be bigger,
1203 * it will be reallocated in mdt_fix_reply.
1204 * 100 stripes is a bit less than 2.5k of data */
1205 #define DEF_REP_MD_SIZE (sizeof(struct lov_mds_md) + \
1206 100 * sizeof(struct lov_ost_data))
1208 #define XATTR_NAME_ACL_ACCESS "system.posix_acl_access"
1209 #define XATTR_NAME_ACL_DEFAULT "system.posix_acl_default"
1210 #define XATTR_USER_PREFIX "user."
1211 #define XATTR_TRUSTED_PREFIX "trusted."
1212 #define XATTR_SECURITY_PREFIX "security."
1213 #define XATTR_ENCRYPTION_PREFIX "encryption."
1215 #define XATTR_NAME_SOM "trusted.som"
1216 #define XATTR_NAME_LOV "trusted.lov"
1217 #define XATTR_NAME_LMA "trusted.lma"
1218 #define XATTR_NAME_LMV "trusted.lmv"
1219 #define XATTR_NAME_DEFAULT_LMV "trusted.dmv"
1220 #define XATTR_NAME_LINK "trusted.link"
1221 #define XATTR_NAME_FID "trusted.fid"
1222 #define XATTR_NAME_VERSION "trusted.version"
1223 #define XATTR_NAME_SOM "trusted.som"
1224 #define XATTR_NAME_HSM "trusted.hsm"
1225 #define XATTR_NAME_LFSCK_BITMAP "trusted.lfsck_bitmap"
1226 #define XATTR_NAME_DUMMY "trusted.dummy"
1227 #define XATTR_NAME_PROJID "trusted.projid"
1228 #define XATTR_NAME_DATAVER "trusted.dataver"
1230 #define LL_XATTR_NAME_ENCRYPTION_CONTEXT_OLD XATTR_SECURITY_PREFIX"c"
1231 #define LL_XATTR_NAME_ENCRYPTION_CONTEXT XATTR_ENCRYPTION_PREFIX"c"
1233 #define XATTR_NAME_LFSCK_NAMESPACE "trusted.lfsck_ns"
1234 #define XATTR_NAME_MAX_LEN 32 /* increase this, if there is longer name. */
1236 #define XATTR_NAME_JOB_DEFAULT "user.job"
1237 /* longest allowed jobid xattr name is "user." + 7 chars + null terminator */
1238 #define XATTR_JOB_MAX_LEN 13
1240 struct lov_mds_md_v3 { /* LOV EA mds/wire data (little-endian) */
1241 __u32 lmm_magic; /* magic number = LOV_MAGIC_V3 */
1242 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
1243 struct ost_id lmm_oi; /* LOV object ID */
1244 __u32 lmm_stripe_size; /* size of stripe in bytes */
1245 /* lmm_stripe_count used to be __u32 */
1246 __u16 lmm_stripe_count; /* num stripes in use for this object */
1247 __u16 lmm_layout_gen; /* layout generation number */
1248 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* must be 32bit aligned */
1249 struct lov_ost_data_v1 lmm_objects[0]; /* per-stripe data */
1252 static inline __u32 lov_mds_md_size(__u16 stripes, __u32 lmm_magic)
1254 if (stripes == (__u16)-1)
1257 if (lmm_magic == LOV_MAGIC_V3)
1258 return sizeof(struct lov_mds_md_v3) +
1259 stripes * sizeof(struct lov_ost_data_v1);
1261 return sizeof(struct lov_mds_md_v1) +
1262 stripes * sizeof(struct lov_ost_data_v1);
1266 lov_mds_md_max_stripe_count(__kernel_size_t buf_size, __u32 lmm_magic)
1268 switch (lmm_magic) {
1269 case LOV_MAGIC_V1: {
1270 struct lov_mds_md_v1 lmm;
1272 if (buf_size < sizeof(lmm))
1275 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1277 case LOV_MAGIC_V3: {
1278 struct lov_mds_md_v3 lmm;
1280 if (buf_size < sizeof(lmm))
1283 return (buf_size - sizeof(lmm)) / sizeof(lmm.lmm_objects[0]);
1290 #define OBD_MD_FLID (0x00000001ULL) /* object ID */
1291 #define OBD_MD_FLATIME (0x00000002ULL) /* access time */
1292 #define OBD_MD_FLMTIME (0x00000004ULL) /* data modification time */
1293 #define OBD_MD_FLCTIME (0x00000008ULL) /* change time */
1294 #define OBD_MD_FLSIZE (0x00000010ULL) /* size */
1295 #define OBD_MD_FLBLOCKS (0x00000020ULL) /* allocated blocks count */
1296 #define OBD_MD_FLBLKSZ (0x00000040ULL) /* block size */
1297 #define OBD_MD_FLMODE (0x00000080ULL) /* access bits (mode & ~S_IFMT) */
1298 #define OBD_MD_FLTYPE (0x00000100ULL) /* object type (mode & S_IFMT) */
1299 #define OBD_MD_FLUID (0x00000200ULL) /* user ID */
1300 #define OBD_MD_FLGID (0x00000400ULL) /* group ID */
1301 #define OBD_MD_FLFLAGS (0x00000800ULL) /* flags word */
1302 #define OBD_MD_DOM_SIZE (0x00001000ULL) /* Data-on-MDT component size */
1303 #define OBD_MD_FLNLINK (0x00002000ULL) /* link count */
1304 #define OBD_MD_FLPARENT (0x00004000ULL) /* parent FID */
1305 #define OBD_MD_LAYOUT_VERSION (0x00008000ULL) /* OST object layout version */
1306 #define OBD_MD_FLRDEV (0x00010000ULL) /* device number */
1307 #define OBD_MD_FLEASIZE (0x00020000ULL) /* extended attribute data */
1308 #define OBD_MD_LINKNAME (0x00040000ULL) /* symbolic link target */
1309 #define OBD_MD_FLHANDLE (0x00080000ULL) /* file/lock handle */
1310 #define OBD_MD_FLCKSUM (0x00100000ULL) /* bulk data checksum */
1311 /* OBD_MD_FLQOS (0x00200000ULL) has never been used */
1312 /* OBD_MD_FLCOOKIE (0x00800000ULL) obsolete in 2.8 */
1313 #define OBD_MD_FLPRJQUOTA (0x00400000ULL) /* over quota flags sent from ost */
1314 #define OBD_MD_FLGROUP (0x01000000ULL) /* group */
1315 #define OBD_MD_FLFID (0x02000000ULL) /* ->ost write inline fid */
1316 /* OBD_MD_FLEPOCH (0x04000000ULL) obsolete 2.7.50 */
1317 /* ->mds if epoch opens or closes */
1318 #define OBD_MD_FLGRANT (0x08000000ULL) /* ost preallocation space grant */
1319 #define OBD_MD_FLDIREA (0x10000000ULL) /* dir's extended attribute data */
1320 #define OBD_MD_FLUSRQUOTA (0x20000000ULL) /* over quota flags sent from ost */
1321 #define OBD_MD_FLGRPQUOTA (0x40000000ULL) /* over quota flags sent from ost */
1322 #define OBD_MD_FLMODEASIZE (0x80000000ULL) /* EA size will be changed */
1324 #define OBD_MD_MDS (0x0000000100000000ULL) /* where an inode lives on */
1325 /* OBD_MD_REINT (0x0000000200000000ULL) obsolete 1.8 */
1326 #define OBD_MD_MEA (0x0000000400000000ULL) /* CMD split EA */
1327 #define OBD_MD_TSTATE (0x0000000800000000ULL) /* transient state field */
1329 #define OBD_MD_FLXATTR (0x0000001000000000ULL) /* xattr */
1330 #define OBD_MD_FLXATTRLS (0x0000002000000000ULL) /* xattr list */
1331 #define OBD_MD_FLXATTRRM (0x0000004000000000ULL) /* xattr remove */
1332 #define OBD_MD_FLACL (0x0000008000000000ULL) /* ACL */
1333 #define OBD_MD_FLAGSTATFS (0x0000010000000000ULL) /* aggregated statfs */
1334 /* OBD_MD_FLMDSCAPA (0x0000020000000000ULL) obsolete 2.7.54 */
1335 /* OBD_MD_FLOSSCAPA (0x0000040000000000ULL) obsolete 2.7.54 */
1336 /* OBD_MD_FLCKSPLIT (0x0000080000000000ULL) obsolete 2.3.58*/
1337 #define OBD_MD_FLCROSSREF (0x0000100000000000ULL) /* Cross-ref case */
1338 #define OBD_MD_FLGETATTRLOCK (0x0000200000000000ULL) /* Get IOEpoch attributes
1339 * under lock; for xattr
1340 * requests means the
1341 * client holds the lock */
1342 #define OBD_MD_FLOBJCOUNT (0x0000400000000000ULL) /* for multiple destroy */
1344 #define OBD_MD_FLDATAVERSION (0x0010000000000000ULL) /* iversion sum */
1345 #define OBD_MD_CLOSE_INTENT_EXECED (0x0020000000000000ULL) /* close intent
1348 #define OBD_MD_DEFAULT_MEA (0x0040000000000000ULL) /* default MEA */
1349 #define OBD_MD_FLOSTLAYOUT (0x0080000000000000ULL) /* contain ost_layout */
1350 #define OBD_MD_FLPROJID (0x0100000000000000ULL) /* project ID */
1351 #define OBD_MD_SECCTX (0x0200000000000000ULL) /* embed security xattr */
1352 #define OBD_MD_FLLAZYSIZE (0x0400000000000000ULL) /* Lazy size */
1353 #define OBD_MD_FLLAZYBLOCKS (0x0800000000000000ULL) /* Lazy blocks */
1354 #define OBD_MD_FLBTIME (0x1000000000000000ULL) /* birth time */
1355 #define OBD_MD_ENCCTX (0x2000000000000000ULL) /* embed encryption ctx */
1356 #define OBD_MD_NAMEHASH (0x4000000000000000ULL) /* use hash instead of name
1357 * in case of encryption
1360 #define OBD_MD_FLALLQUOTA (OBD_MD_FLUSRQUOTA | \
1361 OBD_MD_FLGRPQUOTA | \
1364 #define OBD_MD_FLGETATTR (OBD_MD_FLID | OBD_MD_FLATIME | OBD_MD_FLMTIME | \
1365 OBD_MD_FLCTIME | OBD_MD_FLSIZE | OBD_MD_FLBLKSZ | \
1366 OBD_MD_FLMODE | OBD_MD_FLTYPE | OBD_MD_FLUID | \
1367 OBD_MD_FLGID | OBD_MD_FLFLAGS | OBD_MD_FLNLINK | \
1368 OBD_MD_FLPARENT | OBD_MD_FLRDEV | OBD_MD_FLGROUP | \
1369 OBD_MD_FLPROJID | OBD_MD_FLBTIME)
1371 #define OBD_MD_FLXATTRALL (OBD_MD_FLXATTR | OBD_MD_FLXATTRLS)
1373 /* don't forget obdo_fid which is way down at the bottom so it can
1374 * come after the definition of llog_cookie */
1378 HSS_CLEARMASK = 0x02,
1379 HSS_ARCHIVE_ID = 0x04,
1382 struct hsm_state_set {
1384 __u32 hss_archive_id;
1386 __u64 hss_clearmask;
1389 /* ost_body.data values for OST_BRW */
1391 #define OBD_BRW_READ 0x01
1392 #define OBD_BRW_WRITE 0x02
1393 #define OBD_BRW_RWMASK (OBD_BRW_READ | OBD_BRW_WRITE)
1394 #define OBD_BRW_NDELAY 0x04 /* Non-delay RPC should be issued for
1395 * this page. Non-delay RPCs have bit
1396 * rq_no_delay set. */
1397 #define OBD_BRW_SYNC 0x08 /* this page is a part of synchronous
1398 * transfer and is not accounted in
1400 #define OBD_BRW_CHECK 0x10
1401 #define OBD_BRW_FROM_GRANT 0x20 /* the osc manages this under llite */
1402 #define OBD_BRW_GRANTED 0x40 /* the ost manages this */
1403 /* OBD_BRW_NOCACHE is currently neither set nor tested */
1404 #define OBD_BRW_NOCACHE 0x80 /* this page is a part of non-cached IO */
1405 #define OBD_BRW_NOQUOTA 0x100 /* do not enforce quota */
1406 #define OBD_BRW_SRVLOCK 0x200 /* Client holds no lock over this page */
1407 #define OBD_BRW_ASYNC 0x400 /* Server may delay commit to disk */
1408 #define OBD_BRW_MEMALLOC 0x800 /* Client runs in the "kswapd" context */
1409 #define OBD_BRW_OVER_USRQUOTA 0x1000 /* Running out of user quota */
1410 #define OBD_BRW_OVER_GRPQUOTA 0x2000 /* Running out of group quota */
1411 #define OBD_BRW_SOFT_SYNC 0x4000 /* This flag notifies the server
1412 * that the client is running low on
1413 * space for unstable pages; asking
1414 * it to sync quickly */
1415 #define OBD_BRW_OVER_PRJQUOTA 0x8000 /* Running out of project quota */
1416 #define OBD_BRW_ROOT_PRJQUOTA 0x10000 /* check project quota for root */
1417 #define OBD_BRW_RDMA_ONLY 0x20000 /* RPC contains RDMA-only pages*/
1418 #define OBD_BRW_SYS_RESOURCE 0x40000 /* page has CAP_SYS_RESOURCE */
1419 #define OBD_BRW_COMPRESSED 0x80000 /* data compressed on client */
1421 #define OBD_BRW_OVER_ALLQUOTA (OBD_BRW_OVER_USRQUOTA | \
1422 OBD_BRW_OVER_GRPQUOTA | \
1423 OBD_BRW_OVER_PRJQUOTA)
1425 #define OBD_BRW_DONE 0x40000000UL /*
1426 * osd-ldiskfs inernal,
1427 * IO has been issued before
1429 #define OBD_BRW_LOCAL1 0x80000000UL /*
1430 * osd-ldiskfs internal,
1431 * page mapped to real block
1434 #define OBD_BRW_LOCALS (OBD_BRW_LOCAL1 | OBD_BRW_DONE)
1436 #define OBD_MAX_GRANT 0x7fffffffUL /* Max grant allowed to one client: 2 GiB */
1438 #define OBD_OBJECT_EOF LUSTRE_EOF
1440 #define OST_MIN_PRECREATE 32
1441 #define OST_MAX_PRECREATE 20000
1444 struct ost_id ioo_oid; /* object ID, if multi-obj BRW */
1445 __u32 ioo_max_brw; /* low 16 bits were o_mode before 2.4,
1446 * now (PTLRPC_BULK_OPS_COUNT - 1) in
1447 * high 16 bits in 2.4 and later */
1448 __u32 ioo_bufcnt; /* number of niobufs for this object */
1451 /* NOTE: IOOBJ_MAX_BRW_BITS defines the _offset_ of the max_brw field in
1452 * ioo_max_brw, NOT the maximum number of bits in PTLRPC_BULK_OPS_BITS.
1453 * That said, ioo_max_brw is a 32-bit field so the limit is also 16 bits. */
1454 #define IOOBJ_MAX_BRW_BITS 16
1455 #define ioobj_max_brw_get(ioo) (((ioo)->ioo_max_brw >> IOOBJ_MAX_BRW_BITS) + 1)
1456 #define ioobj_max_brw_set(ioo, num) \
1457 do { (ioo)->ioo_max_brw = ((num) - 1) << IOOBJ_MAX_BRW_BITS; } while (0)
1459 /* multiple of 8 bytes => can array */
1460 struct niobuf_remote {
1466 /* lock value block communicated between the filter and llite */
1468 /* OST_LVB_ERR_INIT is needed because the return code in rc is
1469 * negative, i.e. because ((MASK + rc) & MASK) != MASK. */
1470 #define OST_LVB_ERR_INIT 0xffbadbad80000000ULL
1471 #define OST_LVB_ERR_MASK 0xffbadbad00000000ULL
1472 #define OST_LVB_IS_ERR(blocks) \
1473 ((blocks & OST_LVB_ERR_MASK) == OST_LVB_ERR_MASK)
1474 #define OST_LVB_SET_ERR(blocks, rc) \
1475 do { blocks = OST_LVB_ERR_INIT + rc; } while (0)
1476 #define OST_LVB_GET_ERR(blocks) (int)(blocks - OST_LVB_ERR_INIT)
1499 * lquota data structures
1502 /* The lquota_id structure is an union of all the possible identifier types that
1503 * can be used with quota, this includes:
1506 * - a FID which can be used for per-directory quota in the future */
1508 struct lu_fid qid_fid; /* FID for per-directory quota */
1509 __u64 qid_uid; /* user identifier */
1510 __u64 qid_gid; /* group identifier */
1511 __u64 qid_projid; /* project identifier */
1514 /* quotactl management */
1515 struct obd_quotactl {
1517 __u32 qc_type; /* see Q_* flag below */
1520 struct obd_dqinfo qc_dqinfo;
1521 struct obd_dqblk qc_dqblk;
1525 #define Q_COPY(out, in, member) (out)->member = (in)->member
1528 * - in and out maybe a type of struct if_quotactl or struct obd_quotactl
1529 * - in and out need not be of the same type.
1531 #define __QCTL_COPY(out, in, need_pname) \
1533 Q_COPY(out, in, qc_cmd); \
1534 Q_COPY(out, in, qc_type); \
1535 Q_COPY(out, in, qc_id); \
1536 Q_COPY(out, in, qc_stat); \
1537 Q_COPY(out, in, qc_dqinfo); \
1538 Q_COPY(out, in, qc_dqblk); \
1539 if (need_pname && LUSTRE_Q_CMD_IS_POOL(in->qc_cmd)) { \
1540 size_t len = strnlen(in->qc_poolname, LOV_MAXPOOLNAME); \
1542 memcpy(out->qc_poolname, in->qc_poolname, len); \
1543 out->qc_poolname[len] = '\0'; \
1547 #define QCTL_COPY(out, in) __QCTL_COPY(out, in, true)
1548 #define QCTL_COPY_NO_PNAME(out, in) __QCTL_COPY(out, in, false)
1550 /* Body of quota request used for quota acquire/release RPCs between quota
1551 * master (aka QMT) and slaves (ak QSD). */
1553 struct lu_fid qb_fid; /* FID of global index packing the pool ID
1554 * and type (data or metadata) as well as
1555 * the quota type (user or group). */
1556 union lquota_id qb_id; /* uid or gid or directory FID */
1557 __u32 qb_flags; /* see below */
1559 __u64 qb_count; /* acquire/release count (kbytes/inodes) */
1560 __u64 qb_usage; /* current slave usage (kbytes/inodes) */
1561 __u64 qb_slv_ver; /* slave index file version */
1562 struct lustre_handle qb_lockh; /* per-ID lock handle */
1563 struct lustre_handle qb_glb_lockh; /* global lock handle */
1564 __u64 qb_padding1[4];
1567 /* When the quota_body is used in the reply of quota global intent
1568 * lock (IT_QUOTA_CONN) reply, qb_fid contains slave index file FID. */
1569 #define qb_slv_fid qb_fid
1570 /* qb_usage is the current qunit (in kbytes/inodes) when quota_body is used in
1572 #define qb_qunit qb_usage
1574 #define QUOTA_DQACQ_FL_ACQ 0x1 /* acquire quota */
1575 #define QUOTA_DQACQ_FL_PREACQ 0x2 /* pre-acquire */
1576 #define QUOTA_DQACQ_FL_REL 0x4 /* release quota */
1577 #define QUOTA_DQACQ_FL_REPORT 0x8 /* report usage */
1579 /* Quota types currently supported */
1581 LQUOTA_TYPE_USR = 0x00, /* maps to USRQUOTA */
1582 LQUOTA_TYPE_GRP = 0x01, /* maps to GRPQUOTA */
1583 LQUOTA_TYPE_PRJ = 0x02, /* maps to PRJQUOTA */
1587 /* There are 2 different resource types on which a quota limit can be enforced:
1588 * - inodes on the MDTs
1589 * - blocks on the OSTs */
1591 LQUOTA_RES_MD = 0x01, /* skip 0 to avoid null oid in FID */
1592 LQUOTA_RES_DT = 0x02,
1594 LQUOTA_FIRST_RES = LQUOTA_RES_MD
1596 #define LQUOTA_NR_RES (LQUOTA_LAST_RES - LQUOTA_FIRST_RES + 1)
1599 * Space accounting support
1600 * Format of an accounting record, providing disk usage information for a given
1603 struct lquota_acct_rec { /* 16 bytes */
1604 __u64 bspace; /* current space in use */
1605 __u64 ispace; /* current # inodes in use */
1609 * Global quota index support
1610 * Format of a global record, providing global quota settings for a given quota
1613 struct lquota_glb_rec { /* 32 bytes */
1614 __u64 qbr_hardlimit; /* quota hard limit, in #inodes or kbytes */
1615 __u64 qbr_softlimit; /* quota soft limit, in #inodes or kbytes */
1616 __u64 qbr_time; /* grace time, in seconds */
1617 __u64 qbr_granted; /* how much is granted to slaves, in #inodes or
1622 * Slave index support
1623 * Format of a slave record, recording how much space is granted to a given
1626 struct lquota_slv_rec { /* 8 bytes */
1627 __u64 qsr_granted; /* space granted to the slave for the key=ID,
1628 * in #inodes or kbytes */
1631 /* Data structures associated with the quota locks */
1633 /* Glimpse descriptor used for the index & per-ID quota locks */
1634 struct ldlm_gl_lquota_desc {
1635 union lquota_id gl_id; /* quota ID subject to the glimpse */
1636 __u64 gl_flags; /* see LQUOTA_FL* below */
1637 __u64 gl_ver; /* new index version */
1638 __u64 gl_hardlimit; /* new hardlimit or qunit value */
1639 __u64 gl_softlimit; /* new softlimit */
1643 #define gl_qunit gl_hardlimit /* current qunit value used when
1644 * glimpsing per-ID quota locks */
1646 /* quota glimpse flags */
1647 #define LQUOTA_FL_EDQUOT 0x1 /* user/group out of quota space on QMT */
1649 /* LVB used with quota (global and per-ID) locks */
1651 __u64 lvb_flags; /* see LQUOTA_FL* above */
1652 __u64 lvb_id_may_rel; /* space that might be released later */
1653 __u64 lvb_id_rel; /* space released by the slave for this ID */
1654 __u64 lvb_id_qunit; /* current qunit value */
1658 /* LVB used with global quota lock */
1659 #define lvb_glb_ver lvb_id_may_rel /* current version of the global index */
1667 #define QUOTA_FIRST_OPC QUOTA_DQACQ
1676 MDS_GETATTR_NAME = 34,
1681 MDS_DISCONNECT = 39,
1684 MDS_PIN = 42, /* obsolete, never used in a release */
1685 MDS_UNPIN = 43, /* obsolete, never used in a release */
1687 MDS_DONE_WRITING = 45, /* obsolete since 2.8.0 */
1689 MDS_QUOTACHECK = 47, /* not used since 2.4 */
1692 MDS_SETXATTR = 50, /* obsolete, now it's MDS_REINT op */
1694 MDS_IS_SUBDIR = 52, /* obsolete, never used in a release */
1696 MDS_HSM_STATE_GET = 54,
1697 MDS_HSM_STATE_SET = 55,
1698 MDS_HSM_ACTION = 56,
1699 MDS_HSM_PROGRESS = 57,
1700 MDS_HSM_REQUEST = 58,
1701 MDS_HSM_CT_REGISTER = 59,
1702 MDS_HSM_CT_UNREGISTER = 60,
1703 MDS_SWAP_LAYOUTS = 61,
1709 #define MDS_FIRST_OPC MDS_GETATTR
1712 /* opcodes for object update */
1718 #define OUT_UPDATE_FIRST_OPC OUT_UPDATE
1738 /* the disposition of the intent outlines what was executed */
1739 #define DISP_IT_EXECD 0x00000001
1740 #define DISP_LOOKUP_EXECD 0x00000002
1741 #define DISP_LOOKUP_NEG 0x00000004
1742 #define DISP_LOOKUP_POS 0x00000008
1743 #define DISP_OPEN_CREATE 0x00000010
1744 #define DISP_OPEN_OPEN 0x00000020
1745 #define DISP_ENQ_COMPLETE 0x00400000 /* obsolete and unused */
1746 #define DISP_ENQ_OPEN_REF 0x00800000
1747 #define DISP_ENQ_CREATE_REF 0x01000000
1748 #define DISP_OPEN_LOCK 0x02000000
1749 #define DISP_OPEN_LEASE 0x04000000
1750 #define DISP_OPEN_STRIPE 0x08000000
1751 #define DISP_OPEN_DENY 0x10000000
1753 /* INODE LOCK PARTS */
1754 enum mds_ibits_locks {
1755 MDS_INODELOCK_LOOKUP = 0x000001, /* For namespace, dentry etc. Was
1756 * used to protect permission (mode,
1757 * owner, group, etc) before 2.4. */
1758 MDS_INODELOCK_UPDATE = 0x000002, /* size, links, timestamps */
1759 MDS_INODELOCK_OPEN = 0x000004, /* For opened files */
1760 MDS_INODELOCK_LAYOUT = 0x000008, /* for layout */
1762 /* The PERM bit is added in 2.4, and is used to protect permission
1763 * (mode, owner, group, ACL, etc.) separate from LOOKUP lock.
1764 * For remote directories (in DNE) these locks will be granted by
1765 * different MDTs (different LDLM namespace).
1767 * For local directory, the MDT always grants UPDATE|PERM together.
1768 * For remote directory, master MDT (where remote directory is) grants
1769 * UPDATE|PERM, and remote MDT (where name entry is) grants LOOKUP_LOCK.
1771 MDS_INODELOCK_PERM = 0x000010,
1772 MDS_INODELOCK_XATTR = 0x000020, /* non-permission extended attrs */
1773 MDS_INODELOCK_DOM = 0x000040, /* Data for Data-on-MDT files */
1774 /* Do not forget to increase MDS_INODELOCK_NUMBITS when adding bits */
1776 #define MDS_INODELOCK_NUMBITS 7
1777 /* This FULL lock is useful to take on unlink sort of operations */
1778 #define MDS_INODELOCK_FULL ((1 << MDS_INODELOCK_NUMBITS) - 1)
1779 /* DOM lock shouldn't be canceled early, use this macro for ELC */
1780 #define MDS_INODELOCK_ELC (MDS_INODELOCK_FULL & ~MDS_INODELOCK_DOM)
1782 /* NOTE: until Lustre 1.8.7/2.1.1 the fid_ver() was packed into name[2],
1783 * but was moved into name[1] along with the OID to avoid consuming the
1784 * name[2,3] fields that need to be used for the quota id (also a FID). */
1786 LUSTRE_RES_ID_SEQ_OFF = 0,
1787 LUSTRE_RES_ID_VER_OID_OFF = 1,
1788 LUSTRE_RES_ID_WAS_VER_OFF = 2, /* see note above */
1789 LUSTRE_RES_ID_QUOTA_SEQ_OFF = 2,
1790 LUSTRE_RES_ID_QUOTA_VER_OID_OFF = 3,
1791 LUSTRE_RES_ID_HSH_OFF = 3
1794 #define MDS_STATUS_CONN 1
1795 #define MDS_STATUS_LOV 2
1798 /* these should be identical to their EXT4_*_FL counterparts, they are
1799 * redefined here only to avoid dragging in fs/ext4/ext4.h */
1800 LUSTRE_SYNC_FL = 0x00000008, /* Synchronous updates */
1801 LUSTRE_IMMUTABLE_FL = 0x00000010, /* Immutable file */
1802 LUSTRE_APPEND_FL = 0x00000020, /* file writes may only append */
1803 LUSTRE_NODUMP_FL = 0x00000040, /* do not dump file */
1804 LUSTRE_NOATIME_FL = 0x00000080, /* do not update atime */
1805 LUSTRE_INDEX_FL = 0x00001000, /* hash-indexed directory */
1806 LUSTRE_DIRSYNC_FL = 0x00010000, /* dirsync behaviour (dir only) */
1807 LUSTRE_TOPDIR_FL = 0x00020000, /* Top of directory hierarchies*/
1808 LUSTRE_INLINE_DATA_FL = 0x10000000, /* Inode has inline data. */
1809 LUSTRE_PROJINHERIT_FL = 0x20000000, /* Create with parents projid */
1811 /* These flags will not be identical to any EXT4_*_FL counterparts,
1812 * and only reserved for lustre purpose. Note: these flags might
1813 * be conflict with some of EXT4 flags, so
1814 * 1. these conflict flags needs to be removed when the flag is
1815 * wired by la_flags see osd_attr_get().
1816 * 2. If these flags needs to be stored into inode, they will be
1817 * stored in LMA. see LMAI_XXXX */
1818 LUSTRE_ORPHAN_FL = 0x00002000,
1819 LUSTRE_SET_SYNC_FL = 0x00040000, /* Synchronous setattr on OSTs */
1820 LUSTRE_ENCRYPT_FL = 0x00800000, /* encrypted file */
1822 LUSTRE_LMA_FL_MASKS = LUSTRE_ENCRYPT_FL | LUSTRE_ORPHAN_FL,
1825 #ifndef FS_XFLAG_SYNC
1826 #define FS_XFLAG_SYNC 0x00000020 /* all writes synchronous */
1828 #ifndef FS_XFLAG_NOATIME
1829 #define FS_XFLAG_NOATIME 0x00000040 /* do not update access time */
1831 #ifndef FS_XFLAG_IMMUTABLE
1832 #define FS_XFLAG_IMMUTABLE 0x00000008 /* file cannot be modified */
1834 #ifndef FS_XFLAG_APPEND
1835 #define FS_XFLAG_APPEND 0x00000010 /* all writes append */
1837 #ifndef FS_XFLAG_PROJINHERIT
1838 #define FS_XFLAG_PROJINHERIT 0x00000200 /* create with parents projid */
1841 /* 64 possible states */
1842 enum md_transient_state {
1843 MS_RESTORE = (1 << 0), /* restore is running */
1847 struct lu_fid mbo_fid1;
1848 struct lu_fid mbo_fid2;
1849 struct lustre_handle mbo_open_handle;
1851 __u64 mbo_size; /* Offset, in the case of MDS_READPAGE */
1855 __u64 mbo_blocks; /* XID, in the case of MDS_READPAGE */
1856 __u64 mbo_version; /* was mbo_ioepoch before 2.11 */
1857 __u64 mbo_t_state; /* transient file state defined in
1858 * enum md_transient_state
1859 * was "ino" until 2.4.0 */
1862 __u32 mbo_capability;
1866 __u32 mbo_flags; /* most replies: LUSTRE_*_FL file attributes,
1867 * data_version: OBD_FL_* flags
1870 __u32 mbo_nlink; /* #bytes to read in the case of MDS_READPAGE */
1871 __u32 mbo_layout_gen; /* was "generation" until 2.4.0 */
1873 __u32 mbo_eadatasize;
1875 __u32 mbo_max_mdsize;
1876 __u32 mbo_unused3; /* was max_cookiesize until 2.8 */
1877 __u32 mbo_uid_h; /* high 32-bits of uid, for FUID */
1878 __u32 mbo_gid_h; /* high 32-bits of gid, for FUID */
1880 __u64 mbo_dom_size; /* size of DOM component */
1881 __u64 mbo_dom_blocks; /* blocks consumed by DOM component */
1883 __u64 mbo_padding_9; /* also fix lustre_swab_mdt_body */
1884 __u64 mbo_padding_10;
1887 struct mdt_ioepoch {
1888 struct lustre_handle mio_open_handle;
1889 __u64 mio_unused1; /* was ioepoch */
1890 __u32 mio_unused2; /* was flags */
1894 /* permissions for md_perm.mp_perm */
1896 CFS_SETUID_PERM = 0x01,
1897 CFS_SETGID_PERM = 0x02,
1898 CFS_SETGRP_PERM = 0x04,
1901 struct mdt_rec_setattr {
1911 __u32 sa_padding_1_h;
1912 struct lu_fid sa_fid;
1921 __u32 sa_attr_flags;
1923 __u32 sa_bias; /* some operation flags */
1930 * Attribute flags used in mdt_rec_setattr::sa_valid.
1931 * The kernel's #defines for ATTR_* should not be used over the network
1932 * since the client and MDS may run different kernels (see bug 13828)
1933 * Therefore, we should only use MDS_ATTR_* attributes for sa_valid.
1935 enum mds_attr_flags {
1936 MDS_ATTR_MODE = 0x1ULL, /* = 1 */
1937 MDS_ATTR_UID = 0x2ULL, /* = 2 */
1938 MDS_ATTR_GID = 0x4ULL, /* = 4 */
1939 MDS_ATTR_SIZE = 0x8ULL, /* = 8 */
1940 MDS_ATTR_ATIME = 0x10ULL, /* = 16 */
1941 MDS_ATTR_MTIME = 0x20ULL, /* = 32 */
1942 MDS_ATTR_CTIME = 0x40ULL, /* = 64 */
1943 MDS_ATTR_ATIME_SET = 0x80ULL, /* = 128 */
1944 MDS_ATTR_MTIME_SET = 0x100ULL, /* = 256 */
1945 MDS_ATTR_FORCE = 0x200ULL, /* = 512, change it */
1946 MDS_ATTR_ATTR_FLAG = 0x400ULL, /* = 1024 */
1947 MDS_ATTR_KILL_SUID = 0x800ULL, /* = 2048 */
1948 MDS_ATTR_KILL_SGID = 0x1000ULL, /* = 4096 */
1949 MDS_ATTR_CTIME_SET = 0x2000ULL, /* = 8192 */
1950 MDS_ATTR_FROM_OPEN = 0x4000ULL, /* = 16384, from open O_TRUNC */
1951 MDS_ATTR_BLOCKS = 0x8000ULL, /* = 32768 */
1952 MDS_ATTR_PROJID = 0x10000ULL, /* = 65536 */
1953 MDS_ATTR_LSIZE = 0x20000ULL, /* = 131072 */
1954 MDS_ATTR_LBLOCKS = 0x40000ULL, /* = 262144 */
1955 MDS_ATTR_OVERRIDE = 0x2000000ULL, /* = 33554432 */
1959 /* MDS_CHECK_SPLIT = 1 << 0, obsolete before 2.3.58 */
1960 /* used for remote object getattr/open by name: in the original
1961 * getattr/open request, MDT found the object against name is on another
1962 * MDT, then packed FID and LOOKUP lock in reply and returned -EREMOTE,
1963 * and client knew it's a remote object, then set this flag in
1964 * getattr/open request and sent to the corresponding MDT to finish
1965 * getattr/open, which fetched attributes and UPDATE lock/opened file.
1967 MDS_CROSS_REF = 1 << 1,
1968 /* MDS_VTX_BYPASS = 1 << 2, obsolete since 2.3.54 */
1969 MDS_PERM_BYPASS = 1 << 3,
1970 /* MDS_SOM = 1 << 4, obsolete since 2.8.0 */
1971 MDS_QUOTA_IGNORE = 1 << 5,
1972 /* MDS_CLOSE_CLEANUP = 1 << 6, obsolete since 2.3.51 */
1973 MDS_KEEP_ORPHAN = 1 << 7,
1974 MDS_RECOV_OPEN = 1 << 8,
1975 MDS_DATA_MODIFIED = 1 << 9,
1976 MDS_CREATE_VOLATILE = 1 << 10,
1977 MDS_OWNEROVERRIDE = 1 << 11,
1978 MDS_HSM_RELEASE = 1 << 12,
1979 MDS_CLOSE_MIGRATE = 1 << 13,
1980 MDS_CLOSE_LAYOUT_SWAP = 1 << 14,
1981 MDS_CLOSE_LAYOUT_MERGE = 1 << 15,
1982 MDS_CLOSE_RESYNC_DONE = 1 << 16,
1983 MDS_CLOSE_LAYOUT_SPLIT = 1 << 17,
1984 MDS_TRUNC_KEEP_LEASE = 1 << 18,
1985 MDS_PCC_ATTACH = 1 << 19,
1986 MDS_CLOSE_UPDATE_TIMES = 1 << 20,
1987 /* setstripe create only, don't restripe if target exists */
1988 MDS_SETSTRIPE_CREATE = 1 << 21,
1989 MDS_FID_OP = 1 << 22,
1990 /* migrate dirent only */
1991 MDS_MIGRATE_NSONLY = 1 << 23,
1992 /* create with default LMV from client */
1993 MDS_CREATE_DEFAULT_LMV = 1 << 24,
1996 #define MDS_CLOSE_INTENT (MDS_HSM_RELEASE | MDS_CLOSE_LAYOUT_SWAP | \
1997 MDS_CLOSE_LAYOUT_MERGE | MDS_CLOSE_LAYOUT_SPLIT | \
1998 MDS_CLOSE_RESYNC_DONE)
2000 /* instance of mdt_reint_rec */
2001 struct mdt_rec_create {
2009 __u32 cr_suppgid1_h;
2011 __u32 cr_suppgid2_h;
2012 struct lu_fid cr_fid1;
2013 struct lu_fid cr_fid2;
2014 struct lustre_handle cr_open_handle_old; /* in case of open replay */
2018 __u32 cr_archive_id;
2021 __u64 cr_padding_1; /* rr_blocks */
2024 /* use of helpers set/get_mrc_cr_flags() is needed to access
2025 * 64 bits cr_flags [cr_flags_l, cr_flags_h], this is done to
2026 * extend cr_flags size without breaking 1.8 compat */
2027 __u32 cr_flags_l; /* for use with open, low 32 bits */
2028 __u32 cr_flags_h; /* for use with open, high 32 bits */
2029 __u32 cr_umask; /* umask for create */
2030 __u32 cr_padding_4; /* rr_padding_4 */
2033 /* instance of mdt_reint_rec */
2034 struct mdt_rec_link {
2042 __u32 lk_suppgid1_h;
2044 __u32 lk_suppgid2_h;
2045 struct lu_fid lk_fid1;
2046 struct lu_fid lk_fid2;
2048 __u64 lk_padding_1; /* rr_atime */
2049 __u64 lk_padding_2; /* rr_ctime */
2050 __u64 lk_padding_3; /* rr_size */
2051 __u64 lk_padding_4; /* rr_blocks */
2053 __u32 lk_padding_5; /* rr_mode */
2054 __u32 lk_padding_6; /* rr_flags */
2055 __u32 lk_padding_7; /* rr_padding_2 */
2056 __u32 lk_padding_8; /* rr_padding_3 */
2057 __u32 lk_padding_9; /* rr_padding_4 */
2060 /* instance of mdt_reint_rec */
2061 struct mdt_rec_unlink {
2069 __u32 ul_suppgid1_h;
2071 __u32 ul_suppgid2_h;
2072 struct lu_fid ul_fid1;
2073 struct lu_fid ul_fid2;
2075 __u64 ul_padding_2; /* rr_atime */
2076 __u64 ul_padding_3; /* rr_ctime */
2077 __u64 ul_padding_4; /* rr_size */
2078 __u64 ul_padding_5; /* rr_blocks */
2081 __u32 ul_padding_6; /* rr_flags */
2082 __u32 ul_padding_7; /* rr_padding_2 */
2083 __u32 ul_padding_8; /* rr_padding_3 */
2084 __u32 ul_padding_9; /* rr_padding_4 */
2087 /* instance of mdt_reint_rec */
2088 struct mdt_rec_rename {
2096 __u32 rn_suppgid1_h;
2098 __u32 rn_suppgid2_h;
2099 struct lu_fid rn_fid1;
2100 struct lu_fid rn_fid2;
2102 __u64 rn_padding_1; /* rr_atime */
2103 __u64 rn_padding_2; /* rr_ctime */
2104 __u64 rn_padding_3; /* rr_size */
2105 __u64 rn_padding_4; /* rr_blocks */
2106 __u32 rn_bias; /* some operation flags */
2107 __u32 rn_mode; /* cross-ref rename has mode */
2108 __u32 rn_padding_5; /* rr_flags */
2109 __u32 rn_padding_6; /* rr_padding_2 */
2110 __u32 rn_padding_7; /* rr_padding_3 */
2111 __u32 rn_padding_8; /* rr_padding_4 */
2114 /* instance of mdt_reint_rec */
2115 struct mdt_rec_setxattr {
2123 __u32 sx_suppgid1_h;
2125 __u32 sx_suppgid2_h;
2126 struct lu_fid sx_fid;
2127 __u64 sx_padding_1; /* These three are rr_fid2 */
2132 __u64 sx_padding_5; /* rr_ctime */
2133 __u64 sx_padding_6; /* rr_size */
2134 __u64 sx_padding_7; /* rr_blocks */
2137 __u32 sx_padding_8; /* rr_flags */
2138 __u32 sx_padding_9; /* rr_padding_2 */
2139 __u32 sx_padding_10; /* rr_padding_3 */
2140 __u32 sx_padding_11; /* rr_padding_4 */
2143 /* instance of mdt_reint_rec
2144 * FLR: for file resync MDS_REINT_RESYNC RPC. */
2145 struct mdt_rec_resync {
2153 __u32 rs_suppgid1_h;
2155 __u32 rs_suppgid2_h;
2156 struct lu_fid rs_fid;
2157 __u8 rs_padding0[sizeof(struct lu_fid)];
2158 struct lustre_handle rs_lease_handle; /* rr_mtime */
2159 __s64 rs_padding1; /* rr_atime */
2160 __s64 rs_padding2; /* rr_ctime */
2161 __u64 rs_padding3; /* rr_size */
2162 __u64 rs_padding4; /* rr_blocks */
2164 __u32 rs_padding5; /* rr_mode */
2165 __u32 rs_padding6; /* rr_flags */
2166 __u32 rs_padding7; /* rr_flags_h */
2167 __u32 rs_padding8; /* rr_umask */
2169 __u16 rs_padding9; /* rr_padding_4 */
2173 * mdt_rec_reint is the template for all mdt_reint_xxx structures.
2174 * Do NOT change the size of various members, otherwise the value
2175 * will be broken in lustre_swab_mdt_rec_reint().
2177 * If you add new members in other mdt_reint_xxx structres and need to use the
2178 * rr_padding_x fields, then update lustre_swab_mdt_rec_reint() also.
2180 struct mdt_rec_reint {
2188 __u32 rr_suppgid1_h;
2190 __u32 rr_suppgid2_h;
2191 struct lu_fid rr_fid1;
2192 struct lu_fid rr_fid2;
2204 __u16 rr_padding_4; /* also fix lustre_swab_mdt_rec_reint */
2207 #define LMV_DESC_QOS_MAXAGE_DEFAULT 60 /* Seconds */
2209 /* lmv structures */
2211 __u32 ld_tgt_count; /* how many MDS's */
2212 __u32 ld_active_tgt_count; /* how many active */
2213 __u32 ld_default_stripe_count; /* how many objects are used */
2214 __u32 ld_pattern; /* default hash pattern */
2215 __u64 ld_default_hash_size;
2216 __u64 ld_padding_1; /* also fix lustre_swab_lmv_desc */
2217 __u32 ld_padding_2; /* also fix lustre_swab_lmv_desc */
2218 __u32 ld_qos_maxage; /* in second */
2219 __u32 ld_padding_3; /* also fix lustre_swab_lmv_desc */
2220 __u32 ld_padding_4; /* also fix lustre_swab_lmv_desc */
2221 struct obd_uuid ld_uuid;
2224 /* LMV layout EA, and it will be stored both in master and slave object */
2225 struct lmv_mds_md_v1 {
2227 __u32 lmv_stripe_count;
2228 __u32 lmv_master_mdt_index; /* On master object, it is master
2229 * MDT index, on slave object, it
2230 * is stripe index of the slave obj */
2231 __u32 lmv_hash_type; /* dir stripe policy, i.e. indicate
2232 * which hash function to be used,
2233 * Note: only lower 16 bits is being
2234 * used for now. Higher 16 bits will
2235 * be used to mark the object status,
2236 * for example migrating or dead. */
2237 __u32 lmv_layout_version; /* increased each time layout changed,
2238 * by directory migration, restripe
2240 __u32 lmv_migrate_offset; /* once this is set, it means this
2241 * directory is been migrated, stripes
2242 * before this offset belong to target,
2243 * from this to source. */
2244 __u32 lmv_migrate_hash; /* hash type of source stripes of
2245 * migrating directory */
2248 char lmv_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
2249 struct lu_fid lmv_stripe_fids[0]; /* FIDs for each stripe */
2252 /* stripe count before directory split */
2253 #define lmv_split_offset lmv_migrate_offset
2254 /* stripe count after directory merge */
2255 #define lmv_merge_offset lmv_migrate_offset
2256 /* directory hash type after merge */
2257 #define lmv_merge_hash lmv_migrate_hash
2259 /* foreign LMV EA */
2260 struct lmv_foreign_md {
2261 __u32 lfm_magic; /* magic number = LMV_MAGIC_FOREIGN */
2262 __u32 lfm_length; /* length of lfm_value */
2263 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
2264 __u32 lfm_flags; /* flags, type specific */
2265 char lfm_value[]; /* free format value */
2268 #define LMV_MAGIC_V1 0x0CD20CD0 /* normal stripe lmv magic */
2269 #define LMV_MAGIC LMV_MAGIC_V1
2271 /* #define LMV_USER_MAGIC 0x0CD30CD0 */
2272 #define LMV_MAGIC_STRIPE 0x0CD40CD0 /* magic for dir sub_stripe */
2273 #define LMV_MAGIC_FOREIGN 0x0CD50CD0 /* magic for lmv foreign */
2276 * The FNV-1a hash algorithm is as follows:
2277 * hash = FNV_offset_basis
2278 * for each octet_of_data to be hashed
2279 * hash = hash XOR octet_of_data
2280 * hash = hash × FNV_prime
2282 * http://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function#FNV-1a_hash
2284 * http://www.isthe.com/chongo/tech/comp/fnv/index.html#FNV-reference-source
2285 * FNV_prime is 2^40 + 2^8 + 0xb3 = 0x100000001b3ULL
2287 #define LUSTRE_FNV_1A_64_PRIME 0x100000001b3ULL
2288 #define LUSTRE_FNV_1A_64_OFFSET_BIAS 0xcbf29ce484222325ULL
2289 static inline __u64 lustre_hash_fnv_1a_64(const void *buf, __kernel_size_t size)
2291 __u64 hash = LUSTRE_FNV_1A_64_OFFSET_BIAS;
2292 const unsigned char *p = buf;
2295 for (i = 0; i < size; i++) {
2297 hash *= LUSTRE_FNV_1A_64_PRIME;
2303 /* CRUSH placement group count */
2304 #define LMV_CRUSH_PG_COUNT 4096
2308 struct lmv_mds_md_v1 lmv_md_v1;
2309 struct lmv_user_md lmv_user_md;
2310 struct lmv_foreign_md lmv_foreign_md;
2313 static inline __kernel_ssize_t lmv_mds_md_size(int stripe_count,
2314 unsigned int lmm_magic)
2316 __kernel_ssize_t len = -EINVAL;
2318 switch (lmm_magic) {
2319 case LMV_MAGIC_V1: {
2320 struct lmv_mds_md_v1 *lmm1;
2322 len = sizeof(*lmm1);
2323 len += stripe_count * sizeof(lmm1->lmv_stripe_fids[0]);
2331 static inline int lmv_mds_md_stripe_count_get(const union lmv_mds_md *lmm)
2333 switch (__le32_to_cpu(lmm->lmv_magic)) {
2335 return __le32_to_cpu(lmm->lmv_md_v1.lmv_stripe_count);
2336 case LMV_USER_MAGIC:
2337 return __le32_to_cpu(lmm->lmv_user_md.lum_stripe_count);
2343 static inline int lmv_mds_md_hash_type_get(const union lmv_mds_md *lmm)
2345 switch (__le32_to_cpu(lmm->lmv_magic)) {
2347 return __le32_to_cpu(lmm->lmv_md_v1.lmv_hash_type);
2348 case LMV_USER_MAGIC:
2349 return __le32_to_cpu(lmm->lmv_user_md.lum_hash_type);
2359 FLD_FIRST_OPC = FLD_QUERY
2365 SEQ_FIRST_OPC = SEQ_QUERY
2369 SEQ_ALLOC_SUPER = 0,
2381 LFSCK_NOTIFY = 1101,
2384 LFSCK_FIRST_OPC = LFSCK_NOTIFY
2388 * LOV data structures
2391 #define LOV_MAX_UUID_BUFFER_SIZE 8192
2392 /* The size of the buffer the lov/mdc reserves for the
2393 * array of UUIDs returned by the MDS. With the current
2394 * protocol, this will limit the max number of OSTs per LOV */
2396 #define LOV_DESC_MAGIC 0xB0CCDE5C
2397 #define LOV_DESC_QOS_MAXAGE_DEFAULT 5 /* Seconds */
2398 #define LOV_DESC_STRIPE_SIZE_DEFAULT (1 << LNET_MTU_BITS)
2400 /* LOV settings descriptor (should only contain static info) */
2402 __u32 ld_tgt_count; /* how many OBD's */
2403 __u32 ld_active_tgt_count; /* how many active */
2404 __s32 ld_default_stripe_count; /* how many objects are used */
2405 __u32 ld_pattern; /* default PATTERN_RAID0 */
2406 __u64 ld_default_stripe_size; /* in bytes */
2407 __s64 ld_default_stripe_offset; /* starting OST index */
2408 __u32 ld_padding_0; /* unused */
2409 __u32 ld_qos_maxage; /* in second */
2410 __u32 ld_padding_1; /* also fix lustre_swab_lov_desc */
2411 __u32 ld_padding_2; /* also fix lustre_swab_lov_desc */
2412 struct obd_uuid ld_uuid;
2415 #define ld_magic ld_active_tgt_count /* for swabbing from llogs */
2420 /* opcodes -- MUST be distinct from OST/MDS opcodes */
2425 LDLM_BL_CALLBACK = 104,
2426 LDLM_CP_CALLBACK = 105,
2427 LDLM_GL_CALLBACK = 106,
2428 LDLM_SET_INFO = 107,
2431 #define LDLM_FIRST_OPC LDLM_ENQUEUE
2433 #define RES_NAME_SIZE 4
2434 struct ldlm_res_id {
2435 __u64 name[RES_NAME_SIZE];
2438 #define DLDLMRES "[%#llx:%#llx:%#llx].%#llx"
2439 #define PLDLMRES(res) (unsigned long long)(res)->lr_name.name[0], \
2440 (unsigned long long)(res)->lr_name.name[1], \
2441 (unsigned long long)(res)->lr_name.name[2], \
2442 (unsigned long long)(res)->lr_name.name[3]
2459 #define LCK_MODE_NUM 9
2469 #define LDLM_MIN_TYPE LDLM_PLAIN
2471 struct ldlm_extent {
2477 static inline bool ldlm_extent_equal(const struct ldlm_extent *ex1,
2478 const struct ldlm_extent *ex2)
2480 return ex1->start == ex2->start && ex1->end == ex2->end;
2483 struct ldlm_inodebits {
2486 __u64 try_bits; /* optional bits to try */
2487 __u64 cancel_bits; /* for lock convert */
2491 __u32 li_initiator_id; /* index of MDT that initiated this lock */
2494 struct ldlm_flock_wire {
2502 /* it's important that the fields of the ldlm_extent structure match
2503 * the first fields of the ldlm_flock structure because there is only
2504 * one ldlm_swab routine to process the ldlm_policy_data_t union. if
2505 * this ever changes we will need to swab the union differently based
2506 * on the resource type. */
2508 union ldlm_wire_policy_data {
2509 struct ldlm_extent l_extent;
2510 struct ldlm_flock_wire l_flock;
2511 struct ldlm_inodebits l_inodebits;
2514 struct barrier_lvb {
2520 struct ldlm_gl_barrier_desc {
2526 union ldlm_gl_desc {
2527 struct ldlm_gl_lquota_desc lquota_desc;
2528 struct ldlm_gl_barrier_desc barrier_desc;
2531 enum ldlm_intent_flags {
2532 IT_OPEN = 0x00000001,
2533 IT_CREAT = 0x00000002,
2534 IT_OPEN_CREAT = IT_OPEN | IT_CREAT, /* To allow case label. */
2535 IT_READDIR = 0x00000004, /* Used by mdc, not put on the wire. */
2536 IT_GETATTR = 0x00000008,
2537 IT_LOOKUP = 0x00000010,
2538 /* IT_UNLINK = 0x00000020, Obsolete. */
2539 /* IT_TRUNC = 0x00000040, Obsolete. */
2540 IT_GETXATTR = 0x00000080,
2541 /* IT_EXEC = 0x00000100, Obsolete. */
2542 /* IT_PIN = 0x00000200, Obsolete. */
2543 IT_LAYOUT = 0x00000400,
2544 IT_QUOTA_DQACQ = 0x00000800,
2545 IT_QUOTA_CONN = 0x00001000,
2546 /* IT_SETXATTR = 0x00002000, Obsolete. */
2547 IT_GLIMPSE = 0x00004000,
2548 IT_BRW = 0x00008000,
2551 struct ldlm_intent {
2555 struct ldlm_resource_desc {
2556 enum ldlm_type lr_type;
2557 __u32 lr_pad; /* also fix lustre_swab_ldlm_resource_desc */
2558 struct ldlm_res_id lr_name;
2561 struct ldlm_lock_desc {
2562 struct ldlm_resource_desc l_resource;
2563 enum ldlm_mode l_req_mode;
2564 enum ldlm_mode l_granted_mode;
2565 union ldlm_wire_policy_data l_policy_data;
2568 #define LDLM_LOCKREQ_HANDLES 2
2569 #define LDLM_ENQUEUE_CANCEL_OFF 1
2571 struct ldlm_request {
2572 __u32 lock_flags; /* LDLM_FL_*, see lustre_dlm_flags.h */
2573 __u32 lock_count; /* number of locks in lock_handle[] */
2574 struct ldlm_lock_desc lock_desc;/* lock descriptor */
2575 struct lustre_handle lock_handle[LDLM_LOCKREQ_HANDLES];
2580 __u32 lock_padding; /* also fix lustre_swab_ldlm_reply */
2581 struct ldlm_lock_desc lock_desc;
2582 struct lustre_handle lock_handle;
2583 __u64 lock_policy_res1;
2584 __u64 lock_policy_res2;
2587 #define ldlm_flags_to_wire(flags) ((__u32)(flags))
2588 #define ldlm_flags_from_wire(flags) ((__u64)(flags))
2591 * Opcodes for mountconf (mgs and mgc)
2595 MGS_DISCONNECT = 251,
2596 MGS_EXCEPTION = 252, /* node died, etc. */
2597 MGS_TARGET_REG = 253, /* whenever target starts up */
2598 MGS_TARGET_DEL = 254,
2600 MGS_CONFIG_READ = 256,
2602 MGS_FIRST_OPC = MGS_CONNECT
2605 #if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 18, 53, 0)
2606 #define MGS_PARAM_MAXLEN 1024
2607 #define KEY_SET_INFO "set_info"
2609 struct mgs_send_param {
2610 char mgs_param[MGS_PARAM_MAXLEN];
2614 /* We pass this info to the MGS so it can write config logs */
2615 #define MTI_NAME_MAXLEN 64
2616 #define MTI_PARAM_MAXLEN 4096
2617 #define MTI_NIDS_MAX 32
2619 struct mgs_target_info {
2620 __u32 mti_lustre_ver;
2621 __u32 mti_stripe_index;
2622 __u32 mti_config_ver;
2623 __u32 mti_flags; /* LDD_F_* */
2624 __u32 mti_nid_count;
2625 __u32 mti_instance; /* Running instance of target */
2626 char mti_fsname[MTI_NAME_MAXLEN];
2627 char mti_svname[MTI_NAME_MAXLEN];
2628 char mti_uuid[sizeof(struct obd_uuid)];
2629 __u64 mti_nids[MTI_NIDS_MAX]; /* host nids (lnet_nid_t) */
2630 char mti_params[MTI_PARAM_MAXLEN];
2631 char mti_nidlist[][LNET_NIDSTR_SIZE];
2632 } __attribute__((packed));
2634 struct mgs_nidtbl_entry {
2635 __u64 mne_version; /* table version of this entry */
2636 __u32 mne_instance; /* target instance # */
2637 __u32 mne_index; /* target index */
2638 __u32 mne_length; /* length of this entry - by bytes */
2639 __u8 mne_type; /* target type LDD_F_SV_TYPE_OST/MDT */
2640 __u8 mne_nid_type; /* type of nid(mbz). for ipv6. */
2641 __u8 mne_nid_size; /* size of each NID, by bytes */
2642 __u8 mne_nid_count; /* # of NIDs in buffer */
2644 lnet_nid_t nids[0]; /* variable size buffer for NIDs. */
2649 MGS_CFG_T_CONFIG = 0,
2650 MGS_CFG_T_SPTLRPC = 1,
2651 MGS_CFG_T_RECOVER = 2,
2652 MGS_CFG_T_PARAMS = 3,
2653 MGS_CFG_T_NODEMAP = 4,
2654 MGS_CFG_T_BARRIER = 5,
2658 struct mgs_config_body {
2659 char mcb_name[MTI_NAME_MAXLEN]; /* logname */
2660 __u64 mcb_offset; /* next index of config log to request */
2661 __u16 mcb_type; /* type of log: MGS_CFG_T_[CONFIG|RECOVER] */
2662 __u8 mcb_nm_cur_pass;
2663 __u8 mcb_bits; /* bits unit size of config log */
2664 __u32 mcb_units; /* # of units for bulk transfer */
2667 struct mgs_config_res {
2668 __u64 mcr_offset; /* index of last config log */
2670 __u64 mcr_size; /* size of the log */
2671 __u64 mcr_nm_cur_pass; /* current nodemap config pass */
2675 /* Config marker flags (in config log) */
2676 #define CM_START 0x01
2678 #define CM_SKIP 0x04
2679 #define CM_UPGRADE146 0x08
2680 #define CM_EXCLUDE 0x10
2681 #define CM_START_SKIP (CM_START | CM_SKIP)
2684 __u32 cm_step; /* aka config version */
2686 __u32 cm_vers; /* lustre release version number */
2687 __u32 cm_padding; /* 64 bit align */
2688 __s64 cm_createtime; /*when this record was first created */
2689 __s64 cm_canceltime; /*when this record is no longer valid*/
2690 char cm_tgtname[MTI_NAME_MAXLEN];
2691 char cm_comment[MTI_NAME_MAXLEN];
2695 * Opcodes for multiple servers.
2699 /* OBD_LOG_CANCEL = 401, obsolete since 1.5 */
2700 /* OBD_QC_CALLBACK = 402, obsolete since 2.4 */
2703 OBD_FIRST_OPC = OBD_PING
2707 * llog contexts indices.
2709 * There is compatibility problem with indexes below, they are not
2710 * continuous and must keep their numbers for compatibility needs.
2711 * See LU-5218 for details.
2714 LLOG_CONFIG_ORIG_CTXT = 0,
2715 LLOG_CONFIG_REPL_CTXT = 1,
2716 LLOG_MDS_OST_ORIG_CTXT = 2,
2717 LLOG_MDS_OST_REPL_CTXT = 3, /* kept just to avoid re-assignment */
2718 LLOG_SIZE_ORIG_CTXT = 4,
2719 LLOG_SIZE_REPL_CTXT = 5,
2720 LLOG_TEST_ORIG_CTXT = 8,
2721 LLOG_TEST_REPL_CTXT = 9, /* kept just to avoid re-assignment */
2722 LLOG_CHANGELOG_ORIG_CTXT = 12, /**< changelog generation on mdd */
2723 LLOG_CHANGELOG_REPL_CTXT = 13, /**< changelog access on clients */
2724 /* for multiple changelog consumers */
2725 LLOG_CHANGELOG_USER_ORIG_CTXT = 14,
2726 LLOG_AGENT_ORIG_CTXT = 15, /**< agent requests generation on cdt */
2727 LLOG_UPDATELOG_ORIG_CTXT = 16, /* update log. reserve for the client */
2728 LLOG_UPDATELOG_REPL_CTXT = 17, /* update log. reserve for the client */
2732 /** Identifier for a single log object */
2734 struct ost_id lgl_oi;
2736 } __attribute__((packed));
2738 /** Records written to the CATALOGS list */
2739 #define CATLIST "CATALOGS"
2741 struct llog_logid lci_logid;
2745 } __attribute__((packed));
2747 /* Log data record types - there is no specific reason that these need to
2748 * be related to the RPC opcodes, but no reason not to (may be handy later?)
2750 #define LLOG_OP_MAGIC 0x10600000
2751 #define LLOG_OP_MASK 0xfff00000
2754 LLOG_PAD_MAGIC = LLOG_OP_MAGIC | 0x00000,
2755 OST_SZ_REC = LLOG_OP_MAGIC | 0x00f00,
2756 /* OST_RAID1_REC = LLOG_OP_MAGIC | 0x01000, never used */
2757 MDS_UNLINK_REC = LLOG_OP_MAGIC | 0x10000 | (MDS_REINT << 8) |
2758 REINT_UNLINK, /* obsolete after 2.5.0 */
2759 MDS_UNLINK64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2761 /* MDS_SETATTR_REC = LLOG_OP_MAGIC | 0x12401, obsolete 1.8.0 */
2762 MDS_SETATTR64_REC = LLOG_OP_MAGIC | 0x90000 | (MDS_REINT << 8) |
2764 OBD_CFG_REC = LLOG_OP_MAGIC | 0x20000,
2765 /* PTL_CFG_REC = LLOG_OP_MAGIC | 0x30000, obsolete 1.4.0 */
2766 LLOG_GEN_REC = LLOG_OP_MAGIC | 0x40000,
2767 /* LLOG_JOIN_REC = LLOG_OP_MAGIC | 0x50000, obsolete 1.8.0 */
2768 CHANGELOG_REC = LLOG_OP_MAGIC | 0x60000,
2769 CHANGELOG_USER_REC = LLOG_OP_MAGIC | 0x70000,
2770 CHANGELOG_USER_REC2 = LLOG_OP_MAGIC | 0x70002,
2771 HSM_AGENT_REC = LLOG_OP_MAGIC | 0x80000,
2772 UPDATE_REC = LLOG_OP_MAGIC | 0xa0000, /* Resevered to avoid
2775 LLOG_HDR_MAGIC = LLOG_OP_MAGIC | 0x45539,
2776 LLOG_LOGID_MAGIC = LLOG_OP_MAGIC | 0x4553b,
2779 #define LLOG_REC_HDR_NEEDS_SWABBING(r) \
2780 (((r)->lrh_type & __swab32(LLOG_OP_MASK)) == __swab32(LLOG_OP_MAGIC))
2782 /** Log record header - stored in little endian order.
2783 * Each record must start with this struct, end with a llog_rec_tail,
2784 * and be a multiple of 256 bits in size.
2786 struct llog_rec_hdr {
2791 } __attribute__((packed));
2793 struct llog_rec_tail {
2796 } __attribute__((packed));
2798 /* Where data follow just after header */
2799 #define REC_DATA(ptr) \
2800 ((void *)((char *)ptr + sizeof(struct llog_rec_hdr)))
2802 #define REC_DATA_LEN(rec) \
2803 (rec->lrh_len - sizeof(struct llog_rec_hdr) - \
2804 sizeof(struct llog_rec_tail))
2806 struct llog_logid_rec {
2807 struct llog_rec_hdr lid_hdr;
2808 struct llog_logid lid_id;
2812 struct llog_rec_tail lid_tail;
2813 } __attribute__((packed));
2815 struct llog_unlink_rec {
2816 struct llog_rec_hdr lur_hdr;
2820 struct llog_rec_tail lur_tail;
2821 } __attribute__((packed));
2823 struct llog_unlink64_rec {
2824 struct llog_rec_hdr lur_hdr;
2825 struct lu_fid lur_fid;
2826 __u32 lur_count; /* to destroy the lost precreated */
2830 struct llog_rec_tail lur_tail;
2831 } __attribute__((packed));
2833 struct llog_setattr64_rec {
2834 struct llog_rec_hdr lsr_hdr;
2835 struct ost_id lsr_oi;
2841 struct llog_rec_tail lsr_tail;
2842 } __attribute__((packed));
2844 /* Extended to support project quota */
2845 struct llog_setattr64_rec_v2 {
2846 struct llog_rec_hdr lsr_hdr;
2847 struct ost_id lsr_oi;
2854 __u32 lsr_layout_version;
2857 struct llog_rec_tail lsr_tail;
2858 } __attribute__((packed));
2860 struct llog_size_change_rec {
2861 struct llog_rec_hdr lsc_hdr;
2862 struct ll_fid lsc_fid;
2867 struct llog_rec_tail lsc_tail;
2868 } __attribute__((packed));
2870 #define CHANGELOG_MAGIC 0xca103000
2872 /** \a changelog_rec_type's that can't be masked */
2873 #define CHANGELOG_MINMASK BIT(CL_MARK)
2874 /** bits covering all \a changelog_rec_type's */
2875 #define CHANGELOG_ALLMASK (BIT(CL_LAST) - 1)
2876 /** default \a changelog_rec_type mask. Allow all of them, except
2877 * CL_ATIME since it can really be time consuming, and not necessary
2879 * Remove also CL_OPEN, CL_GETXATTR and CL_DN_OPEN from default list as it can
2880 * be costly and only necessary for audit purpose.
2882 #define CHANGELOG_DEFMASK (CHANGELOG_ALLMASK & \
2883 ~(BIT(CL_ATIME) | BIT(CL_OPEN) | BIT(CL_GETXATTR) | BIT(CL_DN_OPEN)))
2885 /* changelog llog name, needed by client replicators */
2886 #define CHANGELOG_CATALOG "changelog_catalog"
2888 struct changelog_setinfo {
2891 } __attribute__((packed));
2893 /** changelog record */
2894 struct llog_changelog_rec {
2895 struct llog_rec_hdr cr_hdr;
2896 struct changelog_rec cr; /**< Variable length field */
2897 struct llog_rec_tail cr_do_not_use; /**< for_sizeof_only */
2898 } __attribute__((packed));
2900 #define CHANGELOG_USER_PREFIX "cl"
2901 #define CHANGELOG_USER_NAMELEN 16 /* base name including NUL terminator */
2902 #define CHANGELOG_USER_NAMELEN_FULL 30 /* basename plus 'cl$ID-' prefix */
2904 struct llog_changelog_user_rec {
2905 struct llog_rec_hdr cur_hdr;
2907 /* only intended to be used in relative time comparisons to
2908 * detect idle users */
2911 struct llog_rec_tail cur_tail;
2912 } __attribute__((packed));
2914 /* this is twice the size of CHANGELOG_USER_REC */
2915 struct llog_changelog_user_rec2 {
2916 struct llog_rec_hdr cur_hdr;
2918 /* only for use in relative time comparisons to detect idle users */
2923 char cur_name[CHANGELOG_USER_NAMELEN];
2926 struct llog_rec_tail cur_tail;
2927 } __attribute__((packed));
2929 enum agent_req_status {
2937 static inline const char *agent_req_status2name(enum agent_req_status ars)
2955 struct llog_agent_req_rec {
2956 struct llog_rec_hdr arr_hdr; /**< record header */
2957 __u32 arr_status; /**< status of the request */
2959 * agent_req_status */
2960 __u32 arr_archive_id; /**< backend archive number */
2961 __u64 arr_flags; /**< req flags */
2962 __u64 arr_compound_id; /**< compound cookie, ignored */
2963 __u64 arr_req_create; /**< req. creation time */
2964 __u64 arr_req_change; /**< req. status change time */
2965 struct hsm_action_item arr_hai; /**< req. to the agent */
2966 struct llog_rec_tail arr_tail; /**< record tail for_sizezof_only */
2967 } __attribute__((packed));
2969 /* Old llog gen for compatibility */
2973 } __attribute__((packed));
2975 struct llog_gen_rec {
2976 struct llog_rec_hdr lgr_hdr;
2977 struct llog_gen lgr_gen;
2981 struct llog_rec_tail lgr_tail;
2984 /* flags for the logs */
2986 LLOG_F_ZAP_WHEN_EMPTY = 0x1,
2987 LLOG_F_IS_CAT = 0x2,
2988 LLOG_F_IS_PLAIN = 0x4,
2989 LLOG_F_EXT_JOBID = 0x8,
2990 LLOG_F_IS_FIXSIZE = 0x10,
2991 LLOG_F_EXT_EXTRA_FLAGS = 0x20,
2992 LLOG_F_EXT_X_UIDGID = 0x40,
2993 LLOG_F_EXT_X_NID = 0x80,
2994 LLOG_F_EXT_X_OMODE = 0x100,
2995 LLOG_F_EXT_X_XATTR = 0x200,
2996 LLOG_F_RM_ON_ERR = 0x400,
2997 LLOG_F_MAX_AGE = 0x800,
2999 /* Note: Flags covered by LLOG_F_EXT_MASK will be inherited from
3000 * catlog to plain log, so do not add LLOG_F_IS_FIXSIZE here,
3001 * because the catlog record is usually fixed size, but its plain
3002 * log record can be variable */
3003 LLOG_F_EXT_MASK = LLOG_F_EXT_JOBID | LLOG_F_EXT_EXTRA_FLAGS |
3004 LLOG_F_EXT_X_UIDGID | LLOG_F_EXT_X_NID |
3005 LLOG_F_EXT_X_OMODE | LLOG_F_EXT_X_XATTR,
3008 /* On-disk header structure of each log object, stored in little endian order */
3009 #define LLOG_MIN_CHUNK_SIZE 8192
3010 #define LLOG_HEADER_SIZE (96) /* sizeof (llog_log_hdr) + sizeof(llh_tail)
3011 * - sizeof(llh_bitmap) */
3012 #define LLOG_BITMAP_BYTES (LLOG_MIN_CHUNK_SIZE - LLOG_HEADER_SIZE)
3013 #define LLOG_MIN_REC_SIZE (24) /* round(llog_rec_hdr + llog_rec_tail) */
3015 struct llog_log_hdr {
3016 struct llog_rec_hdr llh_hdr;
3017 __s64 llh_timestamp;
3019 __u32 llh_bitmap_offset;
3022 /* for a catalog the first/oldest and still in-use plain slot is just
3023 * next to it. It will serve as the upper limit after Catalog has
3026 struct obd_uuid llh_tgtuuid;
3027 __u32 llh_reserved[LLOG_HEADER_SIZE/sizeof(__u32)-23];
3028 /* These fields must always be at the end of the llog_log_hdr.
3029 * Note: llh_bitmap size is variable because llog chunk size could be
3030 * bigger than LLOG_MIN_CHUNK_SIZE, i.e. sizeof(llog_log_hdr) > 8192
3031 * bytes, and the real size is stored in llh_hdr.lrh_len, which means
3032 * llh_tail should only be refered by LLOG_HDR_TAIL().
3033 * But this structure is also used by client/server llog interface
3034 * (see llog_client.c), it will be kept in its original way to avoid
3035 * compatiblity issue. */
3036 __u32 llh_bitmap[LLOG_BITMAP_BYTES / sizeof(__u32)];
3037 struct llog_rec_tail llh_tail;
3038 } __attribute__((packed));
3039 #undef LLOG_HEADER_SIZE
3040 #undef LLOG_BITMAP_BYTES
3042 #define LLOG_HDR_BITMAP_SIZE(llh) (__u32)((llh->llh_hdr.lrh_len - \
3043 llh->llh_bitmap_offset - \
3044 sizeof(llh->llh_tail)) * 8)
3045 #define LLOG_HDR_BITMAP(llh) (__u32 *)((char *)(llh) + \
3046 (llh)->llh_bitmap_offset)
3047 #define LLOG_HDR_TAIL(llh) ((struct llog_rec_tail *)((char *)llh + \
3048 llh->llh_hdr.lrh_len - \
3049 sizeof(llh->llh_tail)))
3051 /** log cookies are used to reference a specific log file and a record therein,
3052 and pass record offset from llog_process_thread to llog_write */
3053 struct llog_cookie {
3055 struct llog_logid lgc_lgl;
3061 } __attribute__((packed));
3063 /** llog protocol */
3064 enum llogd_rpc_ops {
3065 LLOG_ORIGIN_HANDLE_CREATE = 501,
3066 LLOG_ORIGIN_HANDLE_NEXT_BLOCK = 502,
3067 LLOG_ORIGIN_HANDLE_READ_HEADER = 503,
3068 /* LLOG_ORIGIN_HANDLE_WRITE_REC = 504, Obsolete by 2.1. */
3069 /* LLOG_ORIGIN_HANDLE_CLOSE = 505, Obsolete by 1.8. */
3070 /* LLOG_ORIGIN_CONNECT = 506, Obsolete by 2.4. */
3071 /* LLOG_CATINFO = 507, Obsolete by 2.3. */
3072 LLOG_ORIGIN_HANDLE_PREV_BLOCK = 508,
3073 LLOG_ORIGIN_HANDLE_DESTROY = 509, /* Obsolete by 2.11. */
3075 LLOG_FIRST_OPC = LLOG_ORIGIN_HANDLE_CREATE
3079 struct llog_logid lgd_logid;
3081 __u32 lgd_llh_flags;
3083 __u32 lgd_saved_index;
3085 __u64 lgd_cur_offset;
3086 } __attribute__((packed));
3088 struct llogd_conn_body {
3089 struct llog_gen lgdc_gen;
3090 struct llog_logid lgdc_logid;
3091 __u32 lgdc_ctxt_idx;
3092 } __attribute__((packed));
3094 /* Note: 64-bit types are 64-bit aligned in structure */
3096 __u64 o_valid; /* hot fields in this obdo */
3099 __u64 o_size; /* o_size-o_blocks == ost_lvb */
3103 __u64 o_blocks; /* brw: cli sent cached bytes */
3106 /* 32-bit fields start here: keep an even number of them via padding */
3107 __u32 o_blksize; /* optimal IO blocksize */
3108 __u32 o_mode; /* brw: cli sent cache remain */
3112 __u32 o_nlink; /* brw: checksum */
3114 __u32 o_misc; /* brw: o_dropped */
3116 __u64 o_ioepoch; /* epoch in ost writes */
3117 __u32 o_stripe_idx; /* holds stripe idx */
3119 struct lustre_handle o_handle; /* brw: lock handle to prolong
3121 /* Originally, the field is llog_cookie for destroy with unlink cookie
3122 * from MDS, it is obsolete in 2.8. Then reuse it by client to transfer
3123 * layout and PFL information in IO, setattr RPCs. Since llog_cookie is
3124 * not used on wire any longer, remove it from the obdo, then it can be
3125 * enlarged freely in the further without affect related RPCs.
3127 * sizeof(ost_layout) + sieof(__u32) == sizeof(llog_cookie). */
3128 struct ost_layout o_layout;
3129 __u32 o_layout_version;
3133 __u64 o_data_version; /* getattr: sum of iversion for
3135 * brw: grant space consumed on
3136 * the client for the write */
3138 __u32 o_padding_4; /* also fix
3139 * lustre_swab_obdo() */
3144 #define o_dirty o_blocks
3145 #define o_undirty o_mode
3146 #define o_dropped o_misc
3147 #define o_cksum o_nlink
3148 #define o_grant_used o_data_version
3149 #define o_falloc_mode o_nlink
3151 struct lfsck_request {
3163 __u16 lr_async_windows;
3165 struct lu_fid lr_fid;
3166 struct lu_fid lr_fid2;
3174 struct lfsck_reply {
3181 LE_LASTID_REBUILDING = 1,
3182 LE_LASTID_REBUILT = 2,
3188 /* LE_FID_ACCESSED = 8, moved to lfsck_events_local */
3190 LE_CONDITIONAL_DESTROY = 10,
3191 LE_PAIRS_VERIFY = 11,
3192 LE_SET_LMV_MASTER = 15,
3193 LE_SET_LMV_SLAVE = 16,
3196 enum lfsck_event_flags {
3197 LEF_TO_OST = 0x00000001,
3198 LEF_FROM_OST = 0x00000002,
3199 LEF_SET_LMV_HASH = 0x00000004,
3200 LEF_SET_LMV_ALL = 0x00000008,
3201 LEF_RECHECK_NAME_HASH = 0x00000010,
3202 LEF_QUERY_ALL = 0x00000020,
3205 /* request structure for OST's */
3210 /* Key for FIEMAP to be used in get_info calls */
3211 struct ll_fiemap_info_key {
3213 struct obdo lfik_oa;
3214 struct fiemap lfik_fiemap;
3217 #define IDX_INFO_MAGIC 0x3D37CC37
3219 /* Index file transfer through the network. The server serializes the index into
3220 * a byte stream which is sent to the client via a bulk transfer */
3224 /* reply: see idx_info_flags below */
3227 /* request & reply: number of lu_idxpage (to be) transferred */
3231 /* request: requested attributes passed down to the iterator API */
3234 /* request & reply: index file identifier (FID) */
3235 struct lu_fid ii_fid;
3237 /* reply: version of the index file before starting to walk the index.
3238 * Please note that the version can be modified at any time during the
3242 /* request: hash to start with:
3243 * reply: hash of the first entry of the first lu_idxpage and hash
3244 * of the entry to read next if any */
3245 __u64 ii_hash_start;
3248 /* reply: size of keys in lu_idxpages, minimal one if II_FL_VARKEY is
3252 /* reply: size of records in lu_idxpages, minimal one if II_FL_VARREC
3261 #define II_END_OFF MDS_DIR_END_OFF /* all entries have been read */
3263 /* List of flags used in idx_info::ii_flags */
3264 enum idx_info_flags {
3265 II_FL_NOHASH = 1 << 0, /* client doesn't care about hash value */
3266 II_FL_VARKEY = 1 << 1, /* keys can be of variable size */
3267 II_FL_VARREC = 1 << 2, /* records can be of variable size */
3268 II_FL_NONUNQ = 1 << 3, /* index supports non-unique keys */
3269 II_FL_NOKEY = 1 << 4, /* client doesn't care about key */
3272 #define LIP_MAGIC 0x8A6D6B6C
3274 /* 4KB (= LU_PAGE_SIZE) container gathering key/record pairs */
3276 /* 16-byte header */
3279 __u16 lip_nr; /* number of entries in the container */
3280 __u64 lip_pad0; /* additional padding for future use */
3282 /* key/record pairs are stored in the remaining 4080 bytes.
3283 * depending upon the flags in idx_info::ii_flags, each key/record
3284 * pair might be preceded by:
3286 * - the key size (II_FL_VARKEY is set)
3287 * - the record size (II_FL_VARREC is set)
3289 * For the time being, we only support fixed-size key & record. */
3290 char lip_entries[0];
3293 #define LIP_HDR_SIZE (offsetof(struct lu_idxpage, lip_entries))
3295 /* Gather all possible type associated with a 4KB container */
3297 struct lu_dirpage lp_dir; /* for MDS_READPAGE */
3298 struct lu_idxpage lp_idx; /* for OBD_IDX_READ */
3299 char lp_array[LU_PAGE_SIZE];
3302 /* security opcodes */
3305 SEC_CTX_INIT_CONT = 802,
3308 SEC_FIRST_OPC = SEC_CTX_INIT
3311 /** The link ea holds 1 \a link_ea_entry for each hardlink */
3312 #define LINK_EA_MAGIC 0x11EAF1DFUL
3313 struct link_ea_header {
3316 __u64 leh_len; /* total size */
3317 __u32 leh_overflow_time;
3321 /** Hardlink data is name and parent fid.
3322 * Stored in this crazy struct for maximum packing and endian-neutrality
3324 struct link_ea_entry {
3325 /** __u16 stored big-endian, unaligned */
3326 unsigned char lee_reclen[2];
3327 unsigned char lee_parent_fid[sizeof(struct lu_fid)];
3329 } __attribute__((packed));
3331 /** fid2path request/reply structure */
3332 struct getinfo_fid2path {
3333 struct lu_fid gf_fid;
3339 struct lu_fid gf_root_fid[0];
3341 } __attribute__((packed));
3343 /** path2parent request/reply structures */
3345 struct lu_fid gp_fid; /**< parent FID */
3346 __u32 gp_linkno; /**< hardlink number */
3347 __u32 gp_name_size; /**< size of the name field */
3348 char gp_name[0]; /**< zero-terminated link name */
3349 } __attribute__((packed));
3351 enum layout_intent_opc {
3352 LAYOUT_INTENT_ACCESS = 0, /** generic access */
3353 LAYOUT_INTENT_READ = 1, /** not used */
3354 LAYOUT_INTENT_WRITE = 2, /** write file, for comp layout */
3355 LAYOUT_INTENT_GLIMPSE = 3, /** not used */
3356 LAYOUT_INTENT_TRUNC = 4, /** truncate file, for comp layout */
3357 LAYOUT_INTENT_RELEASE = 5, /** reserved for HSM release */
3358 LAYOUT_INTENT_RESTORE = 6, /** reserved for HSM restore */
3359 LAYOUT_INTENT_PCCRO_SET = 7, /** set read-only layout for PCC */
3360 LAYOUT_INTENT_PCCRO_CLEAR = 8, /** clear read-only layout */
3363 /* enqueue layout lock with intent */
3364 struct layout_intent {
3365 __u32 li_opc; /* intent operation for enqueue, read, write etc */
3367 struct lu_extent li_extent;
3368 } __attribute__((packed));
3371 * On the wire version of hsm_progress structure.
3373 * Contains the userspace hsm_progress and some internal fields.
3375 struct hsm_progress_kernel {
3376 /* Field taken from struct hsm_progress */
3377 struct lu_fid hpk_fid;
3379 struct hsm_extent hpk_extent;
3381 __u16 hpk_errval; /* positive val */
3383 /* Additional fields */
3384 __u64 hpk_data_version;
3386 } __attribute__((packed));
3389 * OUT_UPDATE RPC Format
3391 * During the cross-ref operation, the Master MDT, which the client send the
3392 * request to, will disassembly the operation into object updates, then OSP
3393 * will send these updates to the remote MDT to be executed.
3395 * An UPDATE_OBJ RPC does a list of updates. Each update belongs to an
3396 * operation and does a type of modification to an object.
3404 * update (ub_count-th)
3406 * ub_count must be less than or equal to UPDATE_PER_RPC_MAX.
3411 * rc [+ buffers] (1st)
3412 * rc [+ buffers] (2st)
3414 * rc [+ buffers] (nr_count-th)
3416 * ur_count must be less than or equal to UPDATE_PER_RPC_MAX and should usually
3417 * be equal to ub_count.
3421 * Type of each update, if adding/deleting update, please also update
3422 * update_opcode in lustre/target/out_lib.c.
3434 OUT_INDEX_LOOKUP = 9,
3435 OUT_INDEX_INSERT = 10,
3436 OUT_INDEX_DELETE = 11,
3442 OUT_XATTR_LIST = 17,
3447 UPDATE_FL_OST = 0x00000001, /* op from OST (not MDT) */
3448 UPDATE_FL_SYNC = 0x00000002, /* commit before replying */
3449 UPDATE_FL_COMMITTED = 0x00000004, /* op committed globally */
3450 UPDATE_FL_NOLOG = 0x00000008 /* for idempotent updates */
3453 struct object_update_param {
3454 __u16 oup_len; /* length of this parameter */
3458 } __attribute__((packed));
3461 struct object_update {
3462 __u16 ou_type; /* enum update_type */
3463 __u16 ou_params_count; /* update parameters count */
3464 __u32 ou_result_size; /* how many bytes can return */
3465 __u32 ou_flags; /* enum update_flag */
3466 __u32 ou_padding1; /* padding 1 */
3467 __u64 ou_batchid; /* op transno on master */
3468 struct lu_fid ou_fid; /* object to be updated */
3469 struct object_update_param ou_params[0]; /* update params */
3472 #define UPDATE_REQUEST_MAGIC_V1 0xBDDE0001
3473 #define UPDATE_REQUEST_MAGIC_V2 0xBDDE0002
3474 #define UPDATE_REQUEST_MAGIC UPDATE_REQUEST_MAGIC_V2
3475 /* Hold object_updates sending to the remote OUT in single RPC */
3476 struct object_update_request {
3478 __u16 ourq_count; /* number of ourq_updates[] */
3480 struct object_update ourq_updates[0];
3483 #define OUT_UPDATE_HEADER_MAGIC 0xBDDF0001
3484 #define OUT_UPDATE_MAX_INLINE_SIZE 4096
3485 /* Header for updates request between MDTs */
3486 struct out_update_header {
3489 __u32 ouh_inline_length;
3490 __u32 ouh_reply_size;
3491 __u32 ouh_inline_data[0];
3494 struct out_update_buffer {
3499 /* the result of object update */
3500 struct object_update_result {
3507 #define UPDATE_REPLY_MAGIC_V1 0x00BD0001
3508 #define UPDATE_REPLY_MAGIC_V2 0x00BD0002
3509 #define UPDATE_REPLY_MAGIC UPDATE_REPLY_MAGIC_V2
3510 /* Hold object_update_results being replied from the remote OUT. */
3511 struct object_update_reply {
3518 /* read update result */
3519 struct out_read_reply {
3526 #define BUT_REQUEST_MAGIC 0xBADE0001
3527 /* Hold batched updates sending to the remote target in a single RPC */
3528 struct batch_update_request {
3529 /* Magic number: BUT_REQUEST_MAGIC. */
3531 /* Number of sub requests packed in this batched RPC: burq_reqmsg[]. */
3533 /* Unused padding field. */
3536 * Sub request message array. As message feild buffers for each sub
3537 * request are packed after padded lustre_msg.lm_buflens[] arrary, thus
3538 * it can locate the next request message via the function
3539 * @batch_update_reqmsg_next() in lustre/include/obj_update.h
3541 struct lustre_msg burq_reqmsg[0];
3544 #define BUT_HEADER_MAGIC 0xBADF0001
3545 /* Header for Batched UpdaTes request */
3546 struct but_update_header {
3547 /* Magic number: BUT_HEADER_MAGIC */
3550 * When the total request buffer length is less than MAX_INLINE_SIZE,
3551 * @buh_count is set with 1 and the batched RPC reqeust can be packed
3553 * Otherwise, @buh_count indicates the IO vector count transferring in
3557 /* inline buffer length when the batched RPC can be packed inline. */
3558 __u32 buh_inline_length;
3559 /* The reply buffer size the client prepared. */
3560 __u32 buh_reply_size;
3561 /* Sub request count in this batched RPC. */
3562 __u32 buh_update_count;
3563 /* Unused padding field. */
3565 /* Inline buffer used when the RPC request can be packed inline. */
3566 __u32 buh_inline_data[0];
3569 struct but_update_buffer {
3574 #define BUT_REPLY_MAGIC 0x00AD0001
3575 /* Batched reply received from a remote targer in a batched RPC. */
3576 struct batch_update_reply {
3577 /* Magic number: BUT_REPLY_MAGIC. */
3579 /* Successful returned sub requests. */
3581 /* Unused padding field. */
3584 * Sub reply message array.
3585 * It can locate the next reply message buffer via the function
3586 * @batch_update_repmsg_next() in lustre/include/obj_update.h
3588 struct lustre_msg burp_repmsg[0];
3592 * Batch update opcode.
3594 enum batch_update_cmd {
3597 BUT_FIRST_OPC = BUT_GETATTR,
3600 /** layout swap request structure
3601 * fid1 and fid2 are in mdt_body
3603 struct mdc_swap_layouts {
3605 } __attribute__((packed));
3607 #define INLINE_RESYNC_ARRAY_SIZE 15
3608 struct close_data_resync_done {
3610 __u32 resync_ids_inline[INLINE_RESYNC_ARRAY_SIZE];
3614 struct lustre_handle cd_handle;
3615 struct lu_fid cd_fid;
3616 __u64 cd_data_version;
3618 __u64 cd_reserved[8];
3619 struct close_data_resync_done cd_resync;
3623 __u32 cd_archive_id;
3627 /* Update llog format */
3629 struct lu_fid uop_fid;
3631 __u16 uop_param_count;
3632 __u16 uop_params_off[];
3633 } __attribute__((packed));
3636 struct update_op uops_op[0];
3639 struct update_params {
3640 struct object_update_param up_params[0];
3643 enum update_records_flag {
3644 UPDATE_RECORD_CONTINUE = 1 >> 0,
3647 * This is the update record format used to store the updates in
3648 * disk. All updates of the operation will be stored in ur_ops.
3649 * All of parameters for updates of the operation will be stored
3651 * To save the space of the record, parameters in ur_ops will only
3652 * remember their offset in ur_params, so to avoid storing duplicate
3653 * parameters in ur_params, which can help us save a lot space for
3654 * operation like creating striped directory.
3656 struct update_records {
3657 __u64 ur_master_transno;
3660 /* If the operation includes multiple updates, then ur_index
3661 * means the index of the update inside the whole updates. */
3663 __u32 ur_update_count;
3664 __u32 ur_param_count;
3665 struct update_ops ur_ops;
3666 /* Note ur_ops has a variable size, so comment out
3667 * the following ur_params, in case some use it directly
3668 * update_records->ur_params
3670 * struct update_params ur_params;
3674 struct llog_update_record {
3675 struct llog_rec_hdr lur_hdr;
3676 struct update_records lur_update_rec;
3677 /* Note ur_update_rec has a variable size, so comment out
3678 * the following ur_tail, in case someone use it directly
3680 * struct llog_rec_tail lur_tail;
3684 /* sepol string format is:
3685 * <1-digit for SELinux status>:<policy name>:<policy version>:<policy hash>
3687 /* Max length of the sepol string
3688 * Should be large enough to contain a sha512sum of the policy
3690 #define SELINUX_MODE_LEN 1
3691 #define SELINUX_POLICY_VER_LEN 3 /* 3 chars to leave room for the future */
3692 #define SELINUX_POLICY_HASH_LEN 64
3693 #define LUSTRE_NODEMAP_SEPOL_LENGTH (SELINUX_MODE_LEN + NAME_MAX + \
3694 SELINUX_POLICY_VER_LEN + \
3695 SELINUX_POLICY_HASH_LEN + 3)
3697 /** enums containing the types of ids contained in a nodemap
3698 * kept so other modules (mgs, mdt, etc) can define the type
3702 enum nodemap_id_type {
3708 enum nodemap_tree_type {
3709 NODEMAP_FS_TO_CLIENT,
3710 NODEMAP_CLIENT_TO_FS,
3713 enum nodemap_mapping_modes {
3714 NODEMAP_MAP_BOTH_LEGACY = 0x0, /* for compatibility */
3715 NODEMAP_MAP_UID = 0x01,
3716 NODEMAP_MAP_GID = 0x02,
3717 NODEMAP_MAP_BOTH = 0x03, /* for compatibility */
3718 NODEMAP_MAP_PROJID = 0x04,
3719 NODEMAP_MAP_ALL = NODEMAP_MAP_UID |
3724 enum nodemap_rbac_roles {
3725 NODEMAP_RBAC_FILE_PERMS = 0x00000001,
3726 NODEMAP_RBAC_DNE_OPS = 0x00000002,
3727 NODEMAP_RBAC_QUOTA_OPS = 0x00000004,
3728 NODEMAP_RBAC_BYFID_OPS = 0x00000008,
3729 NODEMAP_RBAC_CHLG_OPS = 0x00000010,
3730 NODEMAP_RBAC_FSCRYPT_ADMIN = 0x00000020,
3731 NODEMAP_RBAC_NONE = (__u32)~(NODEMAP_RBAC_FILE_PERMS |
3732 NODEMAP_RBAC_DNE_OPS |
3733 NODEMAP_RBAC_QUOTA_OPS |
3734 NODEMAP_RBAC_BYFID_OPS |
3735 NODEMAP_RBAC_CHLG_OPS |
3736 NODEMAP_RBAC_FSCRYPT_ADMIN),
3737 NODEMAP_RBAC_ALL = 0xFFFFFFFF, /* future caps ON by default */
3741 * rawobj stuff for GSS
3743 typedef struct netobj_s {
3748 typedef struct rawobj_s {
3755 * following 3 headers must have the same sizes and offsets
3758 __u8 gh_version; /* gss version */
3759 __u8 gh_sp; /* sec part */
3761 __u32 gh_flags; /* wrap flags */
3762 __u32 gh_proc; /* proc */
3763 __u32 gh_seq; /* sequence */
3764 __u32 gh_svc; /* service */
3768 netobj_t gh_handle; /* context handle */
3771 struct gss_rep_header {
3785 struct gss_err_header {
3800 * GSS part of wire context information sent from client, saved and
3801 * used later by server.
3803 struct gss_wire_ctx {
3811 /* This is the lu_ladvise struct which goes out on the wire.
3812 * Corresponds to the userspace arg llapi_lu_ladvise.
3813 * value[1-4] are unspecified fields, used differently by different advices */
3815 __u16 lla_advice; /* advice type */
3816 __u16 lla_value1; /* values for different advice types */
3818 __u64 lla_start; /* first byte of extent for advice */
3819 __u64 lla_end; /* last byte of extent for advice */
3824 /* This is the ladvise_hdr which goes on the wire, corresponds to the userspace
3825 * arg llapi_ladvise_hdr.
3826 * value[1-3] are unspecified fields, used differently by different advices */
3827 struct ladvise_hdr {
3828 __u32 lah_magic; /* LADVISE_MAGIC */
3829 __u32 lah_count; /* number of advices */
3830 __u64 lah_flags; /* from enum ladvise_flag */
3831 __u32 lah_value1; /* unused */
3832 __u32 lah_value2; /* unused */
3833 __u64 lah_value3; /* unused */
3834 struct lu_ladvise lah_advise[0]; /* advices in this header */
3837 #if defined(__cplusplus)