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) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2010, 2017, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/include/lustre/lustre_user.h
34 * Lustre public user-space interface definitions.
37 #ifndef _LUSTRE_USER_H
38 #define _LUSTRE_USER_H
40 /** \defgroup lustreuser lustreuser
45 #include <linux/kernel.h>
46 #include <linux/types.h>
49 # include <linux/fs.h>
50 # include <linux/quota.h>
51 # include <linux/string.h> /* snprintf() */
52 # include <linux/version.h>
53 # include <uapi/linux/lustre/lustre_fiemap.h>
54 #else /* !__KERNEL__ */
57 # include <stdio.h> /* snprintf() */
60 # define NEED_QUOTA_DEFS
61 /* # include <sys/quota.h> - this causes complaints about caddr_t */
62 # include <sys/stat.h>
63 # include <linux/lustre/lustre_fiemap.h>
64 #endif /* __KERNEL__ */
66 /* Handle older distros */
67 #ifndef __ALIGN_KERNEL
68 # define __ALIGN_KERNEL(x, a) __ALIGN_KERNEL_MASK(x, (typeof(x))(a) - 1)
69 # define __ALIGN_KERNEL_MASK(x, mask) (((x) + (mask)) & ~(mask))
72 #if defined(__cplusplus)
77 * This is a temporary solution of adding quota type.
78 * Should be removed as soon as system header is updated.
81 #define LL_MAXQUOTAS 3
83 #define INITQFNAMES { \
84 "user", /* USRQUOTA */ \
85 "group", /* GRPQUOTA */ \
86 "project", /* PRJQUOTA */ \
100 * We need to always use 64bit version because the structure
101 * is shared across entire cluster where 32bit and 64bit machines
104 #if __BITS_PER_LONG != 64 || defined(__ARCH_WANT_STAT64)
105 typedef struct stat64 lstat_t;
106 #define lstat_f lstat64
107 #define fstat_f fstat64
108 #define fstatat_f fstatat64
110 typedef struct stat lstat_t;
111 #define lstat_f lstat
112 #define fstat_f fstat
113 #define fstatat_f fstatat
116 #define HAVE_LOV_USER_MDS_DATA
118 #define LUSTRE_EOF 0xffffffffffffffffULL
121 #define LL_SUPER_MAGIC 0x0BD00BD0
123 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
125 /* FIEMAP flags supported by Lustre */
126 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
128 enum obd_statfs_state {
129 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
130 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
131 OS_STATE_NOPRECREATE = 0x00000004, /**< no object precreation */
132 OS_STATE_UNUSED1 = 0x00000008, /**< obsolete 1.6, was EROFS=30 */
133 OS_STATE_UNUSED2 = 0x00000010, /**< obsolete 1.6, was EROFS=30 */
134 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
135 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
136 OS_STATE_SUM = 0x00000100, /**< aggregated for all tagrets */
137 OS_STATE_NONROT = 0x00000200, /**< non-rotational device */
140 /** filesystem statistics/attributes for target device */
142 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
143 __u64 os_blocks; /* total size in #os_bsize blocks */
144 __u64 os_bfree; /* number of unused blocks */
145 __u64 os_bavail; /* blocks available for allocation */
146 __u64 os_files; /* total number of objects */
147 __u64 os_ffree; /* # objects that could be created */
148 __u8 os_fsid[40]; /* identifier for filesystem */
149 __u32 os_bsize; /* block size in bytes for os_blocks */
150 __u32 os_namelen; /* maximum length of filename in bytes*/
151 __u64 os_maxbytes; /* maximum object size in bytes */
152 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
153 __u32 os_fprecreated; /* objs available now to the caller */
154 /* used in QoS code to find preferred
156 __u32 os_granted; /* space granted for MDS */
157 __u32 os_spare3; /* Unused padding fields. Remember */
158 __u32 os_spare4; /* to fix lustre_swab_obd_statfs() */
166 /** additional filesystem attributes for target device */
167 struct obd_statfs_info {
168 __u32 os_reserved_mb_low; /* reserved mb low */
169 __u32 os_reserved_mb_high; /* reserved mb high */
170 bool os_enable_pre; /* enable pre create logic */
176 * FID is a cluster-wide unique identifier of a file or an object (stripe).
177 * FIDs are never reused.
181 * FID sequence. Sequence is a unit of migration: all files (objects)
182 * with FIDs from a given sequence are stored on the same server.
183 * Lustre should support 2^64 objects, so even if each sequence
184 * has only a single object we can still enumerate 2^64 objects.
187 /* FID number within sequence. */
190 * FID version, used to distinguish different versions (in the sense
191 * of snapshots, etc.) of the same file system object. Not currently
197 static inline bool fid_is_zero(const struct lu_fid *fid)
199 return fid->f_seq == 0 && fid->f_oid == 0;
202 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
203 * MDT-object's FID::f_ver, instead it is the OST-object index in its
204 * parent MDT-object's layout EA. */
205 #define f_stripe_idx f_ver
208 __u32 ol_stripe_size;
209 __u32 ol_stripe_count;
213 } __attribute__((packed));
215 /* The filter_fid structure has changed several times over its lifetime.
216 * For a long time "trusted.fid" held the MDT inode parent FID/IGIF and
217 * stripe_index and the "self FID" (objid/seq) to be able to recover the
218 * OST objects in case of corruption. With the move to 2.4 and OSD-API for
219 * the OST, the "trusted.lma" xattr was added to the OST objects to store
220 * the "self FID" to be consistent with the MDT on-disk format, and the
221 * filter_fid only stored the MDT inode parent FID and stripe index.
223 * In 2.10, the addition of PFL composite layouts required more information
224 * to be stored into the filter_fid in order to be able to identify which
225 * component the OST object belonged. As well, the stripe size may vary
226 * between components, so it was no longer safe to assume the stripe size
227 * or stripe_count of a file. This is also more robust for plain layouts.
229 * For ldiskfs OSTs that were formatted with 256-byte inodes, there is not
230 * enough space to store both the filter_fid and LMA in the inode, so they
231 * are packed into struct lustre_ost_attrs on disk in trusted.lma to avoid
232 * an extra seek for every OST object access.
234 * In 2.11, FLR mirror layouts also need to store the layout version and
235 * range so that writes to old versions of the layout are not allowed.
236 * That ensures that mirrored objects are not modified by evicted clients,
237 * and ensures that the components are correctly marked stale on the MDT.
239 struct filter_fid_18_23 {
240 struct lu_fid ff_parent; /* stripe_idx in f_ver */
245 struct filter_fid_24_29 {
246 struct lu_fid ff_parent; /* stripe_idx in f_ver */
249 struct filter_fid_210 {
250 struct lu_fid ff_parent; /* stripe_idx in f_ver */
251 struct ost_layout ff_layout;
255 struct lu_fid ff_parent; /* stripe_idx in f_ver */
256 struct ost_layout ff_layout;
257 __u32 ff_layout_version;
258 __u32 ff_range; /* range of layout version that
259 * write are allowed */
260 } __attribute__((packed));
262 /* Userspace should treat lu_fid as opaque, and only use the following methods
263 * to print or parse them. Other functions (e.g. compare, swab) could be moved
264 * here from lustre_idl.h if needed. */
268 LMAC_HSM = 0x00000001,
269 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
270 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
271 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
272 * under /O/<seq>/d<x>. */
273 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
274 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
275 LMAC_IDX_BACKUP = 0x00000040, /* Has index backup. */
279 * Masks for all features that should be supported by a Lustre version to
280 * access a specific file.
281 * This information is stored in lustre_mdt_attrs::lma_incompat.
284 LMAI_RELEASED = 0x00000001, /* file is released */
285 LMAI_AGENT = 0x00000002, /* agent inode */
286 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
287 is on the remote MDT */
288 LMAI_STRIPED = 0x00000008, /* striped directory inode */
289 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
290 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
291 LMAI_STRIPED | LMAI_ORPHAN)
296 * Following struct for object attributes, that will be kept inode's EA.
297 * Introduced in 2.0 release (please see b15993, for details)
298 * Added to all objects since Lustre 2.4 as contains self FID
300 struct lustre_mdt_attrs {
302 * Bitfield for supported data in this structure. From enum lma_compat.
303 * lma_self_fid and lma_flags are always available.
307 * Per-file incompat feature list. Lustre version should support all
308 * flags set in this field. The supported feature mask is available in
312 /** FID of this inode */
313 struct lu_fid lma_self_fid;
316 struct lustre_ost_attrs {
317 /* Use lustre_mdt_attrs directly for now, need a common header
318 * structure if want to change lustre_mdt_attrs in future. */
319 struct lustre_mdt_attrs loa_lma;
321 /* Below five elements are for OST-object's PFID EA, the
322 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
323 * and the stripe_index (low 16 bits), the size should not exceed
324 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
325 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
326 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
327 * loa_comp_* elements are valid. */
328 struct lu_fid loa_parent_fid;
329 __u32 loa_stripe_size;
331 __u64 loa_comp_start;
336 * Prior to 2.4, the LMA structure also included SOM attributes which has since
337 * been moved to a dedicated xattr
338 * lma_flags was also removed because of lma_compat/incompat fields.
340 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
342 enum lustre_som_flags {
343 /* Unknow or no SoM data, must get size from OSTs. */
344 SOM_FL_UNKNOWN = 0x0000,
345 /* Known strictly correct, FLR or DoM file (SoM guaranteed). */
346 SOM_FL_STRICT = 0x0001,
347 /* Known stale - was right at some point in the past, but it is
348 * known (or likely) to be incorrect now (e.g. opened for write). */
349 SOM_FL_STALE = 0x0002,
350 /* Approximate, may never have been strictly correct,
351 * need to sync SOM data to achieve eventual consistency. */
352 SOM_FL_LAZY = 0x0004,
355 struct lustre_som_attrs {
357 __u16 lsa_reserved[3];
363 * OST object IDentifier.
371 struct lu_fid oi_fid;
375 #define DOSTID "%#llx:%llu"
376 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
377 ((unsigned long long)ostid_id(oi))
379 struct ll_futimes_3 {
381 __u64 lfu_atime_nsec;
383 __u64 lfu_mtime_nsec;
385 __u64 lfu_ctime_nsec;
389 * Maximum number of mirrors currently implemented.
391 #define LUSTRE_MIRROR_COUNT_MAX 16
393 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
395 LL_LEASE_RDLCK = 0x01,
396 LL_LEASE_WRLCK = 0x02,
397 LL_LEASE_UNLCK = 0x04,
400 enum ll_lease_flags {
401 LL_LEASE_RESYNC = 0x1,
402 LL_LEASE_RESYNC_DONE = 0x2,
403 LL_LEASE_LAYOUT_MERGE = 0x4,
404 LL_LEASE_LAYOUT_SPLIT = 0x8,
405 LL_LEASE_PCC_ATTACH = 0x10,
408 #define IOC_IDS_MAX 4096
409 struct ll_ioc_lease {
416 struct ll_ioc_lease_id {
427 * The ioctl naming rules:
428 * LL_* - works on the currently opened filehandle instead of parent dir
429 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
430 * *_MDC_* - gets/sets data related to MDC
431 * *_LOV_* - gets/sets data related to OSC/LOV
432 * *FILE* - called on parent dir and passes in a filename
433 * *STRIPE* - set/get lov_user_md
434 * *INFO - set/get lov_user_mds_data
436 /* lustre_ioctl.h 101-150 */
437 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
438 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
439 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
440 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
441 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
442 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
443 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
444 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
445 /* LL_IOC_RECREATE_OBJ 157 obsolete */
446 /* LL_IOC_RECREATE_FID 157 obsolete */
447 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
448 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
449 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
450 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
451 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
452 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
453 /* IOC_LOV_GETINFO 165 obsolete */
454 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
455 /* LL_IOC_RMTACL 167 obsolete */
456 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
457 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
458 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
459 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
460 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
461 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
462 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
463 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
464 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
465 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
466 /* lustre_ioctl.h 177-210 */
467 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
468 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
469 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
470 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
471 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
472 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
473 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
474 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
475 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
476 struct lustre_swap_layouts)
477 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
478 struct hsm_current_action)
479 /* lustre_ioctl.h 221-232 */
480 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
481 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
482 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
483 #define LL_IOC_RMFID _IOR('f', 242, struct fid_array)
484 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
485 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
486 #define LL_IOC_GET_LEASE _IO('f', 244)
487 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
488 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
489 #define LL_IOC_MIGRATE _IOR('f', 247, int)
490 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
491 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
492 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
493 #define LL_IOC_HEAT_GET _IOWR('f', 251, struct lu_heat)
494 #define LL_IOC_HEAT_SET _IOW('f', 251, __u64)
495 #define LL_IOC_PCC_DETACH _IOW('f', 252, struct lu_pcc_detach)
496 #define LL_IOC_PCC_DETACH_BY_FID _IOW('f', 252, struct lu_pcc_detach_fid)
497 #define LL_IOC_PCC_STATE _IOR('f', 252, struct lu_pcc_state)
499 #ifndef FS_IOC_FSGETXATTR
501 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
504 __u32 fsx_xflags; /* xflags field value (get/set) */
505 __u32 fsx_extsize; /* extsize field value (get/set)*/
506 __u32 fsx_nextents; /* nextents field value (get) */
507 __u32 fsx_projid; /* project identifier (get/set) */
508 unsigned char fsx_pad[12];
510 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
511 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
513 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
514 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
515 #ifndef FS_XFLAG_PROJINHERIT
516 #define FS_XFLAG_PROJINHERIT 0x00000200
520 #define LL_STATFS_LMV 1
521 #define LL_STATFS_LOV 2
522 #define LL_STATFS_NODELAY 4
524 #define IOC_MDC_TYPE 'i'
525 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
526 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
527 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
528 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
530 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
532 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
533 * files, but are unlikely to be used in practice and are not harmful if
534 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
535 * devices and are safe for use on new files. See LU-4209. */
536 /* To be compatible with old statically linked binary we keep the check for
537 * the older 0100000000 flag. This is already removed upstream. LU-812. */
538 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
540 #define FASYNC 00020000 /* fcntl, for BSD compatibility */
542 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
543 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
544 O_LOV_DELAY_CREATE_MASK)
546 #define LL_FILE_IGNORE_LOCK 0x00000001
547 #define LL_FILE_GROUP_LOCKED 0x00000002
548 #define LL_FILE_READAHEA 0x00000004
549 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
550 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
551 #define LL_FILE_FLOCK_WARNING 0x00000020 /* warned about disabled flock */
553 #define LOV_USER_MAGIC_V1 0x0BD10BD0
554 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
555 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
556 #define LOV_USER_MAGIC_V3 0x0BD30BD0
557 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
558 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
559 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
560 #define LOV_USER_MAGIC_FOREIGN 0x0BD70BD0
561 #define LOV_USER_MAGIC_SEL 0x0BD80BD0
563 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
564 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
565 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
567 #define LOV_PATTERN_NONE 0x000
568 #define LOV_PATTERN_RAID0 0x001
569 #define LOV_PATTERN_RAID1 0x002
570 #define LOV_PATTERN_MDT 0x100
571 #define LOV_PATTERN_OVERSTRIPING 0x200
573 #define LOV_PATTERN_F_MASK 0xffff0000
574 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
575 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
576 #define LOV_PATTERN_DEFAULT 0xffffffff
578 #define LOV_OFFSET_DEFAULT ((__u16)-1)
580 static inline bool lov_pattern_supported(__u32 pattern)
582 return (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_RAID0 ||
583 (pattern & ~LOV_PATTERN_F_RELEASED) ==
584 (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING) ||
585 (pattern & ~LOV_PATTERN_F_RELEASED) == LOV_PATTERN_MDT;
588 /* RELEASED and MDT patterns are not valid in many places, so rather than
589 * having many extra checks on lov_pattern_supported, we have this separate
590 * check for non-released, non-DOM components
592 static inline bool lov_pattern_supported_normal_comp(__u32 pattern)
594 return pattern == LOV_PATTERN_RAID0 ||
595 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_OVERSTRIPING);
599 #define LOV_MAXPOOLNAME 15
600 #define LOV_POOLNAMEF "%.15s"
602 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
603 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
604 #define LOV_MAX_STRIPE_COUNT_OLD 160
605 /* This calculation is crafted so that input of 4096 will result in 160
606 * which in turn is equal to old maximal stripe count.
607 * XXX: In fact this is too simpified for now, what it also need is to get
608 * ea_type argument to clearly know how much space each stripe consumes.
610 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
611 * allocation that is sufficient for the current generation of systems.
613 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
614 #define LOV_MAX_STRIPE_COUNT 2000 /* ~((12 * 4096 - 256) / 24) */
615 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
616 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
618 #define XATTR_LUSTRE_PREFIX "lustre."
619 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
621 #define lov_user_ost_data lov_user_ost_data_v1
622 struct lov_user_ost_data_v1 { /* per-stripe data structure */
623 struct ost_id l_ost_oi; /* OST object ID */
624 __u32 l_ost_gen; /* generation of this OST index */
625 __u32 l_ost_idx; /* OST index in LOV */
626 } __attribute__((packed));
628 #define lov_user_md lov_user_md_v1
629 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
630 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
631 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
632 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
633 __u32 lmm_stripe_size; /* size of stripe in bytes */
634 __u16 lmm_stripe_count; /* num stripes in use for this object */
636 __u16 lmm_stripe_offset; /* starting stripe offset in
637 * lmm_objects, use when writing */
638 __u16 lmm_layout_gen; /* layout generation number
639 * used when reading */
641 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
642 } __attribute__((packed, __may_alias__));
644 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
645 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
646 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
647 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
648 __u32 lmm_stripe_size; /* size of stripe in bytes */
649 __u16 lmm_stripe_count; /* num stripes in use for this object */
651 __u16 lmm_stripe_offset; /* starting stripe offset in
652 * lmm_objects, use when writing */
653 __u16 lmm_layout_gen; /* layout generation number
654 * used when reading */
656 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
657 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
658 } __attribute__((packed));
660 struct lov_foreign_md {
661 __u32 lfm_magic; /* magic number = LOV_MAGIC_FOREIGN */
662 __u32 lfm_length; /* length of lfm_value */
663 __u32 lfm_type; /* type, see LU_FOREIGN_TYPE_ */
664 __u32 lfm_flags; /* flags, type specific */
668 #define foreign_size(lfm) (((struct lov_foreign_md *)lfm)->lfm_length + \
669 offsetof(struct lov_foreign_md, lfm_value))
671 #define foreign_size_le(lfm) \
672 (le32_to_cpu(((struct lov_foreign_md *)lfm)->lfm_length) + \
673 offsetof(struct lov_foreign_md, lfm_value))
676 * The stripe size fields are shared for the extension size storage, however
677 * the extension size is stored in KB, not bytes.
679 #define SEL_UNIT_SIZE 1024llu
686 #define DEXT "[%#llx, %#llx)"
687 #define PEXT(ext) (ext)->e_start, (ext)->e_end
689 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
690 struct lu_extent *e2)
692 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
695 static inline bool lu_extent_is_whole(struct lu_extent *e)
697 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
700 enum lov_comp_md_entry_flags {
701 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
702 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
703 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
704 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
705 LCME_FL_OFFLINE = 0x00000008, /* Not used */
706 LCME_FL_INIT = 0x00000010, /* instantiated */
707 LCME_FL_NOSYNC = 0x00000020, /* FLR: no sync for the mirror */
708 LCME_FL_EXTENSION = 0x00000040, /* extension comp, never init */
709 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
710 * won't be stored on disk
714 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
715 LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
718 /* The component flags can be set by users at creation/modification time. */
719 #define LCME_USER_COMP_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
722 /* The mirror flags can be set by users at creation time. */
723 #define LCME_USER_MIRROR_FLAGS (LCME_FL_PREF_RW)
725 /* The allowed flags obtained from the client at component creation time. */
726 #define LCME_CL_COMP_FLAGS (LCME_USER_MIRROR_FLAGS | LCME_FL_EXTENSION)
728 /* The mirror flags sent by client */
729 #define LCME_MIRROR_FLAGS (LCME_FL_NOSYNC)
731 /* These flags have meaning when set in a default layout and will be inherited
732 * from the default/template layout set on a directory.
734 #define LCME_TEMPLATE_FLAGS (LCME_FL_PREF_RW | LCME_FL_NOSYNC | \
737 /* the highest bit in obdo::o_layout_version is used to mark if the file is
739 #define LU_LAYOUT_RESYNC LCME_FL_NEG
741 /* lcme_id can be specified as certain flags, and the the first
742 * bit of lcme_id is used to indicate that the ID is representing
743 * certain LCME_FL_* but not a real ID. Which implies we can have
744 * at most 31 flags (see LCME_FL_XXX). */
747 LCME_ID_MAX = 0x7FFFFFFF,
748 LCME_ID_ALL = 0xFFFFFFFF,
749 LCME_ID_NOT_ID = LCME_FL_NEG
752 #define LCME_ID_MASK LCME_ID_MAX
754 struct lov_comp_md_entry_v1 {
755 __u32 lcme_id; /* unique id of component */
756 __u32 lcme_flags; /* LCME_FL_XXX */
757 struct lu_extent lcme_extent; /* file extent for component */
758 __u32 lcme_offset; /* offset of component blob,
759 start from lov_comp_md_v1 */
760 __u32 lcme_size; /* size of component blob */
761 __u32 lcme_layout_gen;
762 __u64 lcme_timestamp; /* snapshot time if applicable*/
763 __u32 lcme_padding_1;
764 } __attribute__((packed));
766 #define SEQ_ID_MAX 0x0000FFFF
767 #define SEQ_ID_MASK SEQ_ID_MAX
768 /* bit 30:16 of lcme_id is used to store mirror id */
769 #define MIRROR_ID_MASK 0x7FFF0000
770 #define MIRROR_ID_NEG 0x8000
771 #define MIRROR_ID_SHIFT 16
773 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
775 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
778 static inline __u16 mirror_id_of(__u32 id)
780 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
784 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
786 enum lov_comp_md_flags {
787 /* the least 2 bits are used by FLR to record file state */
790 LCM_FL_WRITE_PENDING = 2,
791 LCM_FL_SYNC_PENDING = 3,
792 LCM_FL_FLR_MASK = 0x3,
795 struct lov_comp_md_v1 {
796 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
797 __u32 lcm_size; /* overall size including this struct */
798 __u32 lcm_layout_gen;
800 __u16 lcm_entry_count;
801 /* lcm_mirror_count stores the number of actual mirrors minus 1,
802 * so that non-flr files will have value 0 meaning 1 mirror. */
803 __u16 lcm_mirror_count;
804 __u16 lcm_padding1[3];
806 struct lov_comp_md_entry_v1 lcm_entries[0];
807 } __attribute__((packed));
809 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
811 if (stripes == (__u16)-1)
814 if (lmm_magic == LOV_USER_MAGIC_V1)
815 return sizeof(struct lov_user_md_v1) +
816 stripes * sizeof(struct lov_user_ost_data_v1);
817 return sizeof(struct lov_user_md_v3) +
818 stripes * sizeof(struct lov_user_ost_data_v1);
821 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
822 * use this. It is unsafe to #define those values in this header as it
823 * is possible the application has already #included <sys/stat.h>. */
824 #ifdef HAVE_LOV_USER_MDS_DATA
825 #define lov_user_mds_data lov_user_mds_data_v1
826 struct lov_user_mds_data_v1 {
827 lstat_t lmd_st; /* MDS stat struct */
828 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
829 } __attribute__((packed));
832 struct lmv_user_mds_data {
833 struct lu_fid lum_fid;
839 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
840 LMV_HASH_TYPE_ALL_CHARS = 1,
841 LMV_HASH_TYPE_FNV_1A_64 = 2,
845 #define LMV_HASH_TYPE_DEFAULT LMV_HASH_TYPE_FNV_1A_64
847 #define LMV_HASH_NAME_ALL_CHARS "all_char"
848 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
850 /* not real hash type, but exposed to user as "space" hash type */
851 #define LMV_HASH_NAME_SPACE "space"
853 /* Right now only the lower part(0-16bits) of lmv_hash_type is being used,
854 * and the higher part will be the flag to indicate the status of object,
855 * for example the object is being migrated. And the hash function
856 * might be interpreted differently with different flags. */
857 #define LMV_HASH_TYPE_MASK 0x0000ffff
859 /* once this is set on a plain directory default layout, newly created
860 * subdirectories will be distributed on all MDTs by space usage.
862 #define LMV_HASH_FLAG_SPACE 0x08000000
864 /* The striped directory has ever lost its master LMV EA, then LFSCK
865 * re-generated it. This flag is used to indicate such case. It is an
867 #define LMV_HASH_FLAG_LOST_LMV 0x10000000
869 #define LMV_HASH_FLAG_BAD_TYPE 0x20000000
870 #define LMV_HASH_FLAG_MIGRATION 0x80000000
872 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
874 struct lustre_foreign_type {
876 const char *lft_name;
880 * LOV/LMV foreign types
882 enum lustre_foreign_types {
883 LU_FOREIGN_TYPE_NONE = 0,
884 LU_FOREIGN_TYPE_DAOS = 0xda05,
885 /* must be the max/last one */
886 LU_FOREIGN_TYPE_UNKNOWN = 0xffffffff,
889 extern struct lustre_foreign_type lu_foreign_types[];
891 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
892 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
893 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
894 #define lmv_user_md lmv_user_md_v1
895 struct lmv_user_md_v1 {
896 __u32 lum_magic; /* must be the first field */
897 __u32 lum_stripe_count; /* dirstripe count */
898 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
899 __u32 lum_hash_type; /* Dir stripe policy */
900 __u32 lum_type; /* LMV type: default */
904 char lum_pool_name[LOV_MAXPOOLNAME + 1];
905 struct lmv_user_mds_data lum_objects[0];
906 } __attribute__((packed));
908 static inline __u32 lmv_foreign_to_md_stripes(__u32 size)
910 if (size <= sizeof(struct lmv_user_md))
913 size -= sizeof(struct lmv_user_md);
914 return (size + sizeof(struct lmv_user_mds_data) - 1) /
915 sizeof(struct lmv_user_mds_data);
919 * NB, historically default layout didn't set type, but use XATTR name to differ
920 * from normal layout, for backward compatibility, define LMV_TYPE_DEFAULT 0x0,
921 * and still use the same method.
924 LMV_TYPE_DEFAULT = 0x0000,
927 static inline int lmv_user_md_size(int stripes, int lmm_magic)
929 int size = sizeof(struct lmv_user_md);
931 if (lmm_magic == LMV_USER_MAGIC_SPECIFIC)
932 size += stripes * sizeof(struct lmv_user_mds_data);
937 struct ll_recreate_obj {
943 __u64 id; /* holds object id */
944 __u32 generation; /* holds object generation */
945 __u32 f_type; /* holds object type or stripe idx when passing it to
946 * OST for saving into EA. */
954 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
955 const struct obd_uuid *u2)
957 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
960 static inline int obd_uuid_empty(struct obd_uuid *uuid)
962 return uuid->uuid[0] == '\0';
965 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
967 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
968 uuid->uuid[sizeof(*uuid) - 1] = '\0';
971 /* For printf's only, make sure uuid is terminated */
972 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
977 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
978 /* Obviously not safe, but for printfs, no real harm done...
979 we're always null-terminated, even in a race. */
980 static char temp[sizeof(*uuid)];
981 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
982 temp[sizeof(*uuid) - 1] = '\0';
985 return (char *)(uuid->uuid);
988 #define LUSTRE_MAXFSNAME 8
990 /* Extract fsname from uuid (or target name) of a target
991 e.g. (myfs-OST0007_UUID -> myfs)
992 see also deuuidify. */
993 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
997 strncpy(buf, uuid, buflen - 1);
998 buf[buflen - 1] = '\0';
999 p = strrchr(buf, '-');
1004 /* printf display format for Lustre FIDs
1005 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
1006 #define FID_NOBRACE_LEN 40
1007 #define FID_LEN (FID_NOBRACE_LEN + 2)
1008 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
1009 #define DFID "[" DFID_NOBRACE "]"
1010 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
1012 /* scanf input parse format for fids in DFID_NOBRACE format
1013 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
1014 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
1015 #define SFID "0x%llx:0x%x:0x%x"
1016 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
1018 /********* Quotas **********/
1020 #define LUSTRE_QUOTABLOCK_BITS 10
1021 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
1023 static inline __u64 lustre_stoqb(size_t space)
1025 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
1028 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
1029 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
1030 #define Q_GETOINFO 0x800102 /* get obd quota info */
1031 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
1032 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
1034 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
1035 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
1036 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
1037 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
1038 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
1039 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
1040 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
1041 /* lustre-specific control commands */
1042 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
1043 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
1044 #define LUSTRE_Q_GETDEFAULT 0x80000d /* get default quota */
1045 #define LUSTRE_Q_SETDEFAULT 0x80000e /* set default quota */
1047 /* In the current Lustre implementation, the grace time is either the time
1048 * or the timestamp to be used after some quota ID exceeds the soft limt,
1049 * 48 bits should be enough, its high 16 bits can be used as quota flags.
1051 #define LQUOTA_GRACE_BITS 48
1052 #define LQUOTA_GRACE_MASK ((1ULL << LQUOTA_GRACE_BITS) - 1)
1053 #define LQUOTA_GRACE_MAX LQUOTA_GRACE_MASK
1054 #define LQUOTA_GRACE(t) (t & LQUOTA_GRACE_MASK)
1055 #define LQUOTA_FLAG(t) (t >> LQUOTA_GRACE_BITS)
1056 #define LQUOTA_GRACE_FLAG(t, f) ((__u64)t | (__u64)f << LQUOTA_GRACE_BITS)
1058 /* different quota flags */
1060 /* the default quota flag, the corresponding quota ID will use the default
1061 * quota setting, the hardlimit and softlimit of its quota record in the global
1062 * quota file will be set to 0, the low 48 bits of the grace will be set to 0
1063 * and high 16 bits will contain this flag (see above comment).
1065 #define LQUOTA_FLAG_DEFAULT 0x0001
1067 #define ALLQUOTA 255 /* set all quota */
1068 static inline const char *qtype_name(int qtype)
1081 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
1082 #define SEPOL_DOWNCALL_MAGIC 0x8b8bb842
1085 #define N_PERMS_MAX 64
1087 struct perm_downcall_data {
1093 struct identity_downcall_data {
1100 struct perm_downcall_data idd_perms[N_PERMS_MAX];
1101 __u32 idd_groups[0];
1104 struct sepol_downcall_data {
1106 time_t sdd_sepol_mtime;
1107 __u16 sdd_sepol_len;
1111 #ifdef NEED_QUOTA_DEFS
1113 #define QIF_BLIMITS 1
1115 #define QIF_ILIMITS 4
1116 #define QIF_INODES 8
1117 #define QIF_BTIME 16
1118 #define QIF_ITIME 32
1119 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
1120 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
1121 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
1122 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
1125 #endif /* !__KERNEL__ */
1127 /* lustre volatile file support
1128 * file name header: ".^L^S^T^R:volatile"
1130 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
1131 #define LUSTRE_VOLATILE_HDR_LEN 14
1133 enum lustre_quota_version {
1137 /* XXX: same as if_dqinfo struct in kernel */
1145 /* XXX: same as if_dqblk struct in kernel, plus one padding */
1147 __u64 dqb_bhardlimit;
1148 __u64 dqb_bsoftlimit;
1150 __u64 dqb_ihardlimit;
1151 __u64 dqb_isoftlimit;
1152 __u64 dqb_curinodes;
1166 struct if_quotactl {
1173 struct obd_dqinfo qc_dqinfo;
1174 struct obd_dqblk qc_dqblk;
1176 struct obd_uuid obd_uuid;
1179 /* swap layout flags */
1180 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
1181 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
1182 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
1183 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
1184 #define SWAP_LAYOUTS_CLOSE (1 << 4)
1186 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
1187 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
1188 struct lustre_swap_layouts {
1196 /** Bit-mask of valid attributes */
1197 /* The LA_* flags are written to disk as part of the ChangeLog records
1198 * so they are part of the on-disk and network protocol, and cannot be changed.
1199 * Only the first 12 bits are currently saved.
1202 LA_ATIME = 1 << 0, /* 0x00001 */
1203 LA_MTIME = 1 << 1, /* 0x00002 */
1204 LA_CTIME = 1 << 2, /* 0x00004 */
1205 LA_SIZE = 1 << 3, /* 0x00008 */
1206 LA_MODE = 1 << 4, /* 0x00010 */
1207 LA_UID = 1 << 5, /* 0x00020 */
1208 LA_GID = 1 << 6, /* 0x00040 */
1209 LA_BLOCKS = 1 << 7, /* 0x00080 */
1210 LA_TYPE = 1 << 8, /* 0x00100 */
1211 LA_FLAGS = 1 << 9, /* 0x00200 */
1212 LA_NLINK = 1 << 10, /* 0x00400 */
1213 LA_RDEV = 1 << 11, /* 0x00800 */
1214 LA_BLKSIZE = 1 << 12, /* 0x01000 */
1215 LA_KILL_SUID = 1 << 13, /* 0x02000 */
1216 LA_KILL_SGID = 1 << 14, /* 0x04000 */
1217 LA_PROJID = 1 << 15, /* 0x08000 */
1218 LA_LAYOUT_VERSION = 1 << 16, /* 0x10000 */
1219 LA_LSIZE = 1 << 17, /* 0x20000 */
1220 LA_LBLOCKS = 1 << 18, /* 0x40000 */
1222 * Attributes must be transmitted to OST objects
1224 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
1227 #define MDS_FMODE_READ 00000001
1228 #define MDS_FMODE_WRITE 00000002
1230 #define MDS_FMODE_CLOSED 00000000
1231 #define MDS_FMODE_EXEC 00000004
1232 /* MDS_FMODE_EPOCH 01000000 obsolete since 2.8.0 */
1233 /* MDS_FMODE_TRUNC 02000000 obsolete since 2.8.0 */
1234 /* MDS_FMODE_SOM 04000000 obsolete since 2.8.0 */
1236 #define MDS_OPEN_CREATED 00000010
1237 /* MDS_OPEN_CROSS 00000020 obsolete in 2.12, internal use only */
1239 #define MDS_OPEN_CREAT 00000100
1240 #define MDS_OPEN_EXCL 00000200
1241 #define MDS_OPEN_TRUNC 00001000
1242 #define MDS_OPEN_APPEND 00002000
1243 #define MDS_OPEN_SYNC 00010000
1244 #define MDS_OPEN_DIRECTORY 00200000
1246 #define MDS_OPEN_BY_FID 040000000 /* open_by_fid for known object */
1247 #define MDS_OPEN_DELAY_CREATE 0100000000 /* delay initial object create */
1248 #define MDS_OPEN_OWNEROVERRIDE 0200000000 /* NFSD rw-reopen ro file for owner */
1249 #define MDS_OPEN_JOIN_FILE 0400000000 /* open for join file.
1250 * We do not support JOIN FILE
1251 * anymore, reserve this flags
1252 * just for preventing such bit
1255 #define MDS_OPEN_LOCK 04000000000 /* This open requires open lock */
1256 #define MDS_OPEN_HAS_EA 010000000000 /* specify object create pattern */
1257 #define MDS_OPEN_HAS_OBJS 020000000000 /* Just set the EA the obj exist */
1258 #define MDS_OPEN_NORESTORE 0100000000000ULL /* Do not restore file at open */
1259 #define MDS_OPEN_NEWSTRIPE 0200000000000ULL /* New stripe needed (restripe or
1261 #define MDS_OPEN_VOLATILE 0400000000000ULL /* File is volatile = created
1263 #define MDS_OPEN_LEASE 01000000000000ULL /* Open the file and grant lease
1264 * delegation, succeed if it's not
1265 * being opened with conflict mode.
1267 #define MDS_OPEN_RELEASE 02000000000000ULL /* Open the file for HSM release */
1269 #define MDS_OPEN_RESYNC 04000000000000ULL /* FLR: file resync */
1270 #define MDS_OPEN_PCC 010000000000000ULL /* PCC: auto RW-PCC cache attach
1271 * for newly created file */
1273 /* lustre internal open flags, which should not be set from user space */
1274 #define MDS_OPEN_FL_INTERNAL (MDS_OPEN_HAS_EA | MDS_OPEN_HAS_OBJS | \
1275 MDS_OPEN_OWNEROVERRIDE | MDS_OPEN_LOCK | \
1276 MDS_OPEN_BY_FID | MDS_OPEN_LEASE | \
1277 MDS_OPEN_RELEASE | MDS_OPEN_RESYNC | \
1281 /********* Changelogs **********/
1282 /** Changelog record types */
1283 enum changelog_rec_type {
1286 CL_CREATE = 1, /* namespace */
1287 CL_MKDIR = 2, /* namespace */
1288 CL_HARDLINK = 3, /* namespace */
1289 CL_SOFTLINK = 4, /* namespace */
1290 CL_MKNOD = 5, /* namespace */
1291 CL_UNLINK = 6, /* namespace */
1292 CL_RMDIR = 7, /* namespace */
1293 CL_RENAME = 8, /* namespace */
1294 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
1295 CL_OPEN = 10, /* not currently used */
1296 CL_CLOSE = 11, /* may be written to log only with mtime change */
1297 CL_LAYOUT = 12, /* file layout/striping modified */
1301 CL_XATTR = CL_SETXATTR, /* Deprecated name */
1302 CL_HSM = 16, /* HSM specific events, see flags */
1303 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1307 CL_FLRW = 21, /* FLR: file was firstly written */
1308 CL_RESYNC = 22, /* FLR: file was resync-ed */
1310 CL_DN_OPEN = 24, /* denied open */
1314 static inline const char *changelog_type2str(int type) {
1315 static const char *changelog_str[] = {
1316 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1317 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1318 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1319 "FLRW", "RESYNC","GXATR", "NOPEN",
1322 if (type >= 0 && type < CL_LAST)
1323 return changelog_str[type];
1327 /* 12 bits of per-record data can be stored in the bottom of the flags */
1328 #define CLF_FLAGSHIFT 12
1329 enum changelog_rec_flags {
1330 CLF_VERSION = 0x1000,
1331 CLF_RENAME = 0x2000,
1333 CLF_EXTRA_FLAGS = 0x8000,
1334 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID |
1336 CLF_FLAGMASK = (1U << CLF_FLAGSHIFT) - 1,
1337 CLF_VERMASK = ~CLF_FLAGMASK,
1341 /* Anything under the flagmask may be per-type (if desired) */
1342 /* Flags for unlink */
1343 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1344 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1345 /* HSM cleaning needed */
1346 /* Flags for rename */
1347 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1349 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1350 * has an archive in backend */
1353 /* 12b used (from high weight to low weight):
1358 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1359 #define CLF_HSM_ERR_H 6
1360 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1361 #define CLF_HSM_EVENT_H 9
1362 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1363 #define CLF_HSM_FLAG_H 11
1364 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1365 #define CLF_HSM_SPARE_H 15
1366 #define CLF_HSM_LAST 15
1368 /* Remove bits higher than _h, then extract the value
1369 * between _h and _l by shifting lower weigth to bit 0. */
1370 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1371 >> (CLF_HSM_LAST - _h + _l))
1373 #define CLF_HSM_SUCCESS 0x00
1374 #define CLF_HSM_MAXERROR 0x7E
1375 #define CLF_HSM_ERROVERFLOW 0x7F
1377 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1379 /* 3 bits field => 8 values allowed */
1391 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1393 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1397 static inline void hsm_set_cl_event(enum changelog_rec_flags *clf_flags,
1400 *clf_flags = (enum changelog_rec_flags)
1401 (*clf_flags | (he << CLF_HSM_EVENT_L));
1404 static inline __u16 hsm_get_cl_flags(enum changelog_rec_flags clf_flags)
1406 return CLF_GET_BITS(clf_flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1409 static inline void hsm_set_cl_flags(enum changelog_rec_flags *clf_flags,
1412 *clf_flags = (enum changelog_rec_flags)
1413 (*clf_flags | (bits << CLF_HSM_FLAG_L));
1416 static inline int hsm_get_cl_error(enum changelog_rec_flags clf_flags)
1418 return CLF_GET_BITS(clf_flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1421 static inline void hsm_set_cl_error(enum changelog_rec_flags *clf_flags,
1424 *clf_flags = (enum changelog_rec_flags)
1425 (*clf_flags | (error << CLF_HSM_ERR_L));
1428 enum changelog_rec_extra_flags {
1430 CLFE_UIDGID = 0x0001,
1433 CLFE_XATTR = 0x0008,
1434 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID | CLFE_OPEN | CLFE_XATTR
1437 enum changelog_send_flag {
1438 /* Not yet implemented */
1439 CHANGELOG_FLAG_FOLLOW = 0x01,
1440 /* Blocking IO makes sense in case of slow user parsing of the records,
1441 * but it also prevents us from cleaning up if the records are not
1443 CHANGELOG_FLAG_BLOCK = 0x02,
1444 /* Pack jobid into the changelog records if available. */
1445 CHANGELOG_FLAG_JOBID = 0x04,
1446 /* Pack additional flag bits into the changelog record */
1447 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1450 enum changelog_send_extra_flag {
1451 /* Pack uid/gid into the changelog record */
1452 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1453 /* Pack nid into the changelog record */
1454 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1455 /* Pack open mode into the changelog record */
1456 CHANGELOG_EXTRA_FLAG_OMODE = 0x04,
1457 /* Pack xattr name into the changelog record */
1458 CHANGELOG_EXTRA_FLAG_XATTR = 0x08,
1461 #define CR_MAXSIZE __ALIGN_KERNEL(2 * NAME_MAX + 2 + \
1462 changelog_rec_offset(CLF_SUPPORTED, \
1465 /* 31 usable bytes string + null terminator. */
1466 #define LUSTRE_JOBID_SIZE 32
1468 /* This is the minimal changelog record. It can contain extensions
1469 * such as rename fields or process jobid. Its exact content is described
1470 * by the cr_flags and cr_extra_flags.
1472 * Extensions are packed in the same order as their corresponding flags,
1473 * then in the same order as their corresponding extra flags.
1475 struct changelog_rec {
1477 __u16 cr_flags; /**< \a changelog_rec_flags */
1478 __u32 cr_type; /**< \a changelog_rec_type */
1479 __u64 cr_index; /**< changelog record number */
1480 __u64 cr_prev; /**< last index for this target fid */
1483 struct lu_fid cr_tfid; /**< target fid */
1484 __u32 cr_markerflags; /**< CL_MARK flags */
1486 struct lu_fid cr_pfid; /**< parent fid */
1489 /* Changelog extension for RENAME. */
1490 struct changelog_ext_rename {
1491 struct lu_fid cr_sfid; /**< source fid, or zero */
1492 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1495 /* Changelog extension to include JOBID. */
1496 struct changelog_ext_jobid {
1497 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1500 /* Changelog extension to include additional flags. */
1501 struct changelog_ext_extra_flags {
1502 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1505 /* Changelog extra extension to include UID/GID. */
1506 struct changelog_ext_uidgid {
1511 /* Changelog extra extension to include NID. */
1512 struct changelog_ext_nid {
1513 /* have __u64 instead of lnet_nid_t type for use by client api */
1515 /* for use when IPv6 support is added */
1520 /* Changelog extra extension to include low 32 bits of MDS_OPEN_* flags. */
1521 struct changelog_ext_openmode {
1525 /* Changelog extra extension to include xattr */
1526 struct changelog_ext_xattr {
1527 char cr_xattr[XATTR_NAME_MAX + 1]; /**< zero-terminated string. */
1530 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1531 const struct changelog_rec *rec);
1533 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1534 enum changelog_rec_extra_flags cref)
1536 size_t size = sizeof(struct changelog_rec);
1538 if (crf & CLF_RENAME)
1539 size += sizeof(struct changelog_ext_rename);
1541 if (crf & CLF_JOBID)
1542 size += sizeof(struct changelog_ext_jobid);
1544 if (crf & CLF_EXTRA_FLAGS) {
1545 size += sizeof(struct changelog_ext_extra_flags);
1546 if (cref & CLFE_UIDGID)
1547 size += sizeof(struct changelog_ext_uidgid);
1548 if (cref & CLFE_NID)
1549 size += sizeof(struct changelog_ext_nid);
1550 if (cref & CLFE_OPEN)
1551 size += sizeof(struct changelog_ext_openmode);
1552 if (cref & CLFE_XATTR)
1553 size += sizeof(struct changelog_ext_xattr);
1559 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1561 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1563 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1564 cref = (enum changelog_rec_extra_flags)
1565 changelog_rec_extra_flags(rec)->cr_extra_flags;
1567 return changelog_rec_offset(
1568 (enum changelog_rec_flags)rec->cr_flags, cref);
1571 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1573 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1577 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1579 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1580 (rec->cr_flags & CLF_VERSION);
1582 return (struct changelog_ext_rename *)((char *)rec +
1583 changelog_rec_offset(crf,
1587 /* The jobid follows the rename extension, if present */
1589 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1591 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1592 (rec->cr_flags & (CLF_VERSION | CLF_RENAME));
1594 return (struct changelog_ext_jobid *)((char *)rec +
1595 changelog_rec_offset(crf,
1599 /* The additional flags follow the rename and jobid extensions, if present */
1601 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1602 const struct changelog_rec *rec)
1604 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1605 (rec->cr_flags & (CLF_VERSION | CLF_RENAME | CLF_JOBID));
1607 return (struct changelog_ext_extra_flags *)((char *)rec +
1608 changelog_rec_offset(crf,
1612 /* The uid/gid is the first extra extension */
1614 struct changelog_ext_uidgid *changelog_rec_uidgid(
1615 const struct changelog_rec *rec)
1617 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1619 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1621 return (struct changelog_ext_uidgid *)((char *)rec +
1622 changelog_rec_offset(crf,
1626 /* The nid is the second extra extension */
1628 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1630 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1632 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1633 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1635 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1636 cref = (enum changelog_rec_extra_flags)
1637 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1640 return (struct changelog_ext_nid *)((char *)rec +
1641 changelog_rec_offset(crf, cref));
1644 /* The OPEN mode is the third extra extension */
1646 struct changelog_ext_openmode *changelog_rec_openmode(
1647 const struct changelog_rec *rec)
1649 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1651 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1652 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1654 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1655 cref = (enum changelog_rec_extra_flags)
1656 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1657 (CLFE_UIDGID | CLFE_NID));
1660 return (struct changelog_ext_openmode *)((char *)rec +
1661 changelog_rec_offset(crf, cref));
1664 /* The xattr name is the fourth extra extension */
1666 struct changelog_ext_xattr *changelog_rec_xattr(
1667 const struct changelog_rec *rec)
1669 enum changelog_rec_flags crf = (enum changelog_rec_flags)
1671 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS));
1672 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1674 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1675 cref = (enum changelog_rec_extra_flags)
1676 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1677 (CLFE_UIDGID | CLFE_NID | CLFE_OPEN));
1679 return (struct changelog_ext_xattr *)((char *)rec +
1680 changelog_rec_offset(crf, cref));
1683 /* The name follows the rename, jobid and extra flags extns, if present */
1684 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1686 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1688 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1689 cref = (enum changelog_rec_extra_flags)
1690 changelog_rec_extra_flags(rec)->cr_extra_flags;
1692 return (char *)rec + changelog_rec_offset(
1693 (enum changelog_rec_flags)(rec->cr_flags & CLF_SUPPORTED),
1694 (enum changelog_rec_extra_flags)(cref & CLFE_SUPPORTED));
1697 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1699 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1702 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1704 char *cr_name = changelog_rec_name(rec);
1706 return cr_name + strlen(cr_name) + 1;
1710 * Remap a record to the desired format as specified by the crf flags.
1711 * The record must be big enough to contain the final remapped version.
1712 * Superfluous extension fields are removed and missing ones are added
1713 * and zeroed. The flags of the record are updated accordingly.
1715 * The jobid and rename extensions can be added to a record, to match the
1716 * format an application expects, typically. In this case, the newly added
1717 * fields will be zeroed.
1718 * The Jobid field can be removed, to guarantee compatibility with older
1719 * clients that don't expect this field in the records they process.
1721 * The following assumptions are being made:
1722 * - CLF_RENAME will not be removed
1723 * - CLF_JOBID will not be added without CLF_RENAME being added too
1724 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1726 * @param[in,out] rec The record to remap.
1727 * @param[in] crf_wanted Flags describing the desired extensions.
1728 * @param[in] cref_want Flags describing the desired extra extensions.
1730 static inline void changelog_remap_rec(struct changelog_rec *rec,
1731 enum changelog_rec_flags crf_wanted,
1732 enum changelog_rec_extra_flags cref_want)
1734 char *xattr_mov = NULL;
1735 char *omd_mov = NULL;
1736 char *nid_mov = NULL;
1737 char *uidgid_mov = NULL;
1741 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1743 crf_wanted = (enum changelog_rec_flags)
1744 (crf_wanted & CLF_SUPPORTED);
1745 cref_want = (enum changelog_rec_extra_flags)
1746 (cref_want & CLFE_SUPPORTED);
1748 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1749 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1750 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1751 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1757 /* First move the variable-length name field */
1758 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1759 changelog_rec_name(rec), rec->cr_namelen);
1761 /* Locations of extensions in the remapped record */
1762 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1763 xattr_mov = (char *)rec +
1764 changelog_rec_offset(
1765 (enum changelog_rec_flags)
1766 (crf_wanted & CLF_SUPPORTED),
1767 (enum changelog_rec_extra_flags)
1768 (cref_want & ~CLFE_XATTR));
1769 omd_mov = (char *)rec +
1770 changelog_rec_offset(
1771 (enum changelog_rec_flags)
1772 (crf_wanted & CLF_SUPPORTED),
1773 (enum changelog_rec_extra_flags)
1774 (cref_want & ~(CLFE_OPEN | CLFE_XATTR)));
1775 nid_mov = (char *)rec +
1776 changelog_rec_offset(
1777 (enum changelog_rec_flags)
1778 (crf_wanted & CLF_SUPPORTED),
1779 (enum changelog_rec_extra_flags)
1781 ~(CLFE_NID | CLFE_OPEN | CLFE_XATTR)));
1782 uidgid_mov = (char *)rec +
1783 changelog_rec_offset(
1784 (enum changelog_rec_flags)
1785 (crf_wanted & CLF_SUPPORTED),
1786 (enum changelog_rec_extra_flags)
1787 (cref_want & ~(CLFE_UIDGID |
1791 cref = (enum changelog_rec_extra_flags)
1792 changelog_rec_extra_flags(rec)->cr_extra_flags;
1795 ef_mov = (char *)rec +
1796 changelog_rec_offset(
1797 (enum changelog_rec_flags)
1798 (crf_wanted & ~CLF_EXTRA_FLAGS), CLFE_INVALID);
1799 jid_mov = (char *)rec +
1800 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
1801 ~(CLF_EXTRA_FLAGS | CLF_JOBID)),
1803 rnm_mov = (char *)rec +
1804 changelog_rec_offset((enum changelog_rec_flags)(crf_wanted &
1810 /* Move the extension fields to the desired positions */
1811 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1812 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1813 if ((cref_want & CLFE_XATTR) && (cref & CLFE_XATTR))
1814 memmove(xattr_mov, changelog_rec_xattr(rec),
1815 sizeof(struct changelog_ext_xattr));
1817 if ((cref_want & CLFE_OPEN) && (cref & CLFE_OPEN))
1818 memmove(omd_mov, changelog_rec_openmode(rec),
1819 sizeof(struct changelog_ext_openmode));
1821 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1822 memmove(nid_mov, changelog_rec_nid(rec),
1823 sizeof(struct changelog_ext_nid));
1825 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1826 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1827 sizeof(struct changelog_ext_uidgid));
1829 memmove(ef_mov, changelog_rec_extra_flags(rec),
1830 sizeof(struct changelog_ext_extra_flags));
1833 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1834 memmove(jid_mov, changelog_rec_jobid(rec),
1835 sizeof(struct changelog_ext_jobid));
1837 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1838 memmove(rnm_mov, changelog_rec_rename(rec),
1839 sizeof(struct changelog_ext_rename));
1841 /* Clear newly added fields */
1842 if (xattr_mov && (cref_want & CLFE_XATTR) &&
1843 !(cref & CLFE_XATTR))
1844 memset(xattr_mov, 0, sizeof(struct changelog_ext_xattr));
1846 if (omd_mov && (cref_want & CLFE_OPEN) &&
1847 !(cref & CLFE_OPEN))
1848 memset(omd_mov, 0, sizeof(struct changelog_ext_openmode));
1850 if (nid_mov && (cref_want & CLFE_NID) &&
1852 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1854 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1855 !(cref & CLFE_UIDGID))
1856 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1858 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1859 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1860 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1862 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1863 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1865 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1866 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1868 /* Update the record's flags accordingly */
1869 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1870 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1871 changelog_rec_extra_flags(rec)->cr_extra_flags =
1872 changelog_rec_extra_flags(rec)->cr_extra_flags |
1876 enum changelog_message_type {
1877 CL_RECORD = 10, /* message is a changelog_rec */
1878 CL_EOF = 11, /* at end of current changelog */
1881 /********* Misc **********/
1883 struct ioc_data_version {
1885 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1886 __u32 idv_flags; /* enum ioc_data_version_flags */
1889 enum ioc_data_version_flags {
1890 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1891 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1895 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1898 #define dot_lustre_name ".lustre"
1901 /********* HSM **********/
1903 /** HSM per-file state
1904 * See HSM_FLAGS below.
1907 HS_NONE = 0x00000000,
1908 HS_EXISTS = 0x00000001,
1909 HS_DIRTY = 0x00000002,
1910 HS_RELEASED = 0x00000004,
1911 HS_ARCHIVED = 0x00000008,
1912 HS_NORELEASE = 0x00000010,
1913 HS_NOARCHIVE = 0x00000020,
1914 HS_LOST = 0x00000040,
1917 /* HSM user-setable flags. */
1918 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1920 /* Other HSM flags. */
1921 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1924 * All HSM-related possible flags that could be applied to a file.
1925 * This should be kept in sync with hsm_states.
1927 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1930 * HSM request progress state
1932 enum hsm_progress_states {
1939 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1942 case HPS_WAITING: return "waiting";
1943 case HPS_RUNNING: return "running";
1944 case HPS_DONE: return "done";
1945 default: return "unknown";
1952 } __attribute__((packed));
1955 * Current HSM states of a Lustre file.
1957 * This structure purpose is to be sent to user-space mainly. It describes the
1958 * current HSM flags and in-progress action.
1960 struct hsm_user_state {
1961 /** Current HSM states, from enum hsm_states. */
1963 __u32 hus_archive_id;
1964 /** The current undergoing action, if there is one */
1965 __u32 hus_in_progress_state;
1966 __u32 hus_in_progress_action;
1967 struct hsm_extent hus_in_progress_location;
1968 char hus_extended_info[];
1971 struct hsm_state_set_ioc {
1972 struct lu_fid hssi_fid;
1974 __u64 hssi_clearmask;
1978 * This structure describes the current in-progress action for a file.
1979 * it is retuned to user space and send over the wire
1981 struct hsm_current_action {
1982 /** The current undergoing action, if there is one */
1983 /* state is one of hsm_progress_states */
1985 /* action is one of hsm_user_action */
1987 struct hsm_extent hca_location;
1990 /***** HSM user requests ******/
1991 /* User-generated (lfs/ioctl) request types */
1992 enum hsm_user_action {
1993 HUA_NONE = 1, /* no action (noop) */
1994 HUA_ARCHIVE = 10, /* copy to hsm */
1995 HUA_RESTORE = 11, /* prestage */
1996 HUA_RELEASE = 12, /* drop ost objects */
1997 HUA_REMOVE = 13, /* remove from archive */
1998 HUA_CANCEL = 14 /* cancel a request */
2001 static inline const char *hsm_user_action2name(enum hsm_user_action a)
2004 case HUA_NONE: return "NOOP";
2005 case HUA_ARCHIVE: return "ARCHIVE";
2006 case HUA_RESTORE: return "RESTORE";
2007 case HUA_RELEASE: return "RELEASE";
2008 case HUA_REMOVE: return "REMOVE";
2009 case HUA_CANCEL: return "CANCEL";
2010 default: return "UNKNOWN";
2015 * List of hr_flags (bit field)
2017 #define HSM_FORCE_ACTION 0x0001
2018 /* used by CT, cannot be set by user */
2019 #define HSM_GHOST_COPY 0x0002
2022 * Contains all the fixed part of struct hsm_user_request.
2025 struct hsm_request {
2026 __u32 hr_action; /* enum hsm_user_action */
2027 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
2028 __u64 hr_flags; /* request flags */
2029 __u32 hr_itemcount; /* item count in hur_user_item vector */
2033 struct hsm_user_item {
2034 struct lu_fid hui_fid;
2035 struct hsm_extent hui_extent;
2036 } __attribute__((packed));
2038 struct hsm_user_request {
2039 struct hsm_request hur_request;
2040 struct hsm_user_item hur_user_item[0];
2041 /* extra data blob at end of struct (after all
2042 * hur_user_items), only use helpers to access it
2044 } __attribute__((packed));
2046 /** Return pointer to data field in a hsm user request */
2047 static inline void *hur_data(struct hsm_user_request *hur)
2049 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
2053 * Compute the current length of the provided hsm_user_request. This returns -1
2054 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
2056 * return -1 on bounds check error.
2058 static inline ssize_t hur_len(struct hsm_user_request *hur)
2062 /* can't overflow a __u64 since hr_itemcount is only __u32 */
2063 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
2064 (__u64)hur->hur_request.hr_itemcount *
2065 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
2067 if ((ssize_t)size < 0)
2073 /****** HSM RPCs to copytool *****/
2074 /* Message types the copytool may receive */
2075 enum hsm_message_type {
2076 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
2079 /* Actions the copytool may be instructed to take for a given action_item */
2080 enum hsm_copytool_action {
2081 HSMA_NONE = 10, /* no action */
2082 HSMA_ARCHIVE = 20, /* arbitrary offset */
2088 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
2091 case HSMA_NONE: return "NOOP";
2092 case HSMA_ARCHIVE: return "ARCHIVE";
2093 case HSMA_RESTORE: return "RESTORE";
2094 case HSMA_REMOVE: return "REMOVE";
2095 case HSMA_CANCEL: return "CANCEL";
2096 default: return "UNKNOWN";
2100 /* Copytool item action description */
2101 struct hsm_action_item {
2102 __u32 hai_len; /* valid size of this struct */
2103 __u32 hai_action; /* hsm_copytool_action, but use known size */
2104 struct lu_fid hai_fid; /* Lustre FID to operate on */
2105 struct lu_fid hai_dfid; /* fid used for data access */
2106 struct hsm_extent hai_extent; /* byte range to operate on */
2107 __u64 hai_cookie; /* action cookie from coordinator */
2108 __u64 hai_gid; /* grouplock id */
2109 char hai_data[0]; /* variable length */
2110 } __attribute__((packed));
2113 * helper function which print in hexa the first bytes of
2116 * \param hai [IN] record to print
2117 * \param buffer [IN,OUT] buffer to write the hex string to
2118 * \param len [IN] max buffer length
2122 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
2123 char *buffer, size_t len)
2130 data_len = hai->hai_len - sizeof(*hai);
2131 for (i = 0; (i < data_len) && (len > 2); i++) {
2132 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
2142 /* Copytool action list */
2143 #define HAL_VERSION 1
2144 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
2145 struct hsm_action_list {
2147 __u32 hal_count; /* number of hai's to follow */
2148 __u64 hal_compound_id; /* returned by coordinator, ignored */
2150 __u32 hal_archive_id; /* which archive backend */
2152 char hal_fsname[0]; /* null-terminated */
2153 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
2154 boundaries. See hai_zero */
2155 } __attribute__((packed));
2157 /* Return pointer to first hai in action list */
2158 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
2160 size_t offset = __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2162 return (struct hsm_action_item *)(hal->hal_fsname + offset);
2165 /* Return pointer to next hai */
2166 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
2168 size_t offset = __ALIGN_KERNEL(hai->hai_len, 8);
2170 return (struct hsm_action_item *)((char *)hai + offset);
2173 /* Return size of an hsm_action_list */
2174 static inline size_t hal_size(struct hsm_action_list *hal)
2178 struct hsm_action_item *hai;
2180 sz = sizeof(*hal) + __ALIGN_KERNEL(strlen(hal->hal_fsname) + 1, 8);
2181 hai = hai_first(hal);
2182 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
2183 sz += __ALIGN_KERNEL(hai->hai_len, 8);
2189 * describe the attributes to be set on imported file
2191 struct hsm_user_import {
2200 __u32 hui_archive_id;
2203 /* Copytool progress reporting */
2204 #define HP_FLAG_COMPLETED 0x01
2205 #define HP_FLAG_RETRY 0x02
2207 struct hsm_progress {
2208 struct lu_fid hp_fid;
2210 struct hsm_extent hp_extent;
2212 __u16 hp_errval; /* positive val */
2217 __u64 hc_data_version;
2219 __u16 hc_errval; /* positive val */
2221 struct hsm_action_item hc_hai;
2225 enum llapi_json_types {
2226 LLAPI_JSON_INTEGER = 1,
2232 struct llapi_json_item {
2241 struct llapi_json_item *lji_next;
2244 struct llapi_json_item_list {
2245 int ljil_item_count;
2246 struct llapi_json_item *ljil_items;
2249 enum lu_ladvise_type {
2250 LU_LADVISE_INVALID = 0,
2251 LU_LADVISE_WILLREAD = 1,
2252 LU_LADVISE_DONTNEED = 2,
2253 LU_LADVISE_LOCKNOEXPAND = 3,
2254 LU_LADVISE_LOCKAHEAD = 4,
2258 #define LU_LADVISE_NAMES { \
2259 [LU_LADVISE_WILLREAD] = "willread", \
2260 [LU_LADVISE_DONTNEED] = "dontneed", \
2261 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
2262 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
2265 /* This is the userspace argument for ladvise. It is currently the same as
2266 * what goes on the wire (struct lu_ladvise), but is defined separately as we
2267 * may need info which is only used locally. */
2268 struct llapi_lu_ladvise {
2269 __u16 lla_advice; /* advice type */
2270 __u16 lla_value1; /* values for different advice types */
2272 __u64 lla_start; /* first byte of extent for advice */
2273 __u64 lla_end; /* last byte of extent for advice */
2279 LF_ASYNC = 0x00000001,
2280 LF_UNSET = 0x00000002,
2283 #define LADVISE_MAGIC 0x1ADF1CE0
2284 /* Masks of valid flags for each advice */
2285 #define LF_LOCKNOEXPAND_MASK LF_UNSET
2286 /* Flags valid for all advices not explicitly specified */
2287 #define LF_DEFAULT_MASK LF_ASYNC
2289 #define LF_MASK (LF_ASYNC | LF_UNSET)
2291 #define lla_lockahead_mode lla_value1
2292 #define lla_peradvice_flags lla_value2
2293 #define lla_lockahead_result lla_value3
2295 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
2296 * is used on the wire. It is defined separately as we may need info which is
2297 * only used locally. */
2298 struct llapi_ladvise_hdr {
2299 __u32 lah_magic; /* LADVISE_MAGIC */
2300 __u32 lah_count; /* number of advices */
2301 __u64 lah_flags; /* from enum ladvise_flag */
2302 __u32 lah_value1; /* unused */
2303 __u32 lah_value2; /* unused */
2304 __u64 lah_value3; /* unused */
2305 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
2308 #define LAH_COUNT_MAX (1024)
2312 SK_CRYPT_INVALID = -1,
2314 SK_CRYPT_AES256_CTR = 1,
2318 SK_HMAC_INVALID = -1,
2324 struct sk_crypt_type {
2325 const char *sct_name;
2329 struct sk_hmac_type {
2330 const char *sht_name;
2334 enum lock_mode_user {
2340 #define LOCK_MODE_NAMES { \
2341 [MODE_READ_USER] = "READ",\
2342 [MODE_WRITE_USER] = "WRITE"\
2345 enum lockahead_results {
2346 LLA_RESULT_SENT = 0,
2347 LLA_RESULT_DIFFERENT,
2351 enum lu_heat_flag_bit {
2352 LU_HEAT_FLAG_BIT_INVALID = 0,
2353 LU_HEAT_FLAG_BIT_OFF,
2354 LU_HEAT_FLAG_BIT_CLEAR,
2358 LU_HEAT_FLAG_OFF = 1ULL << LU_HEAT_FLAG_BIT_OFF,
2359 LU_HEAT_FLAG_CLEAR = 1ULL << LU_HEAT_FLAG_BIT_CLEAR,
2362 enum obd_heat_type {
2363 OBD_HEAT_READSAMPLE = 0,
2364 OBD_HEAT_WRITESAMPLE = 1,
2365 OBD_HEAT_READBYTE = 2,
2366 OBD_HEAT_WRITEBYTE = 3,
2370 #define LU_HEAT_NAMES { \
2371 [OBD_HEAT_READSAMPLE] = "readsample", \
2372 [OBD_HEAT_WRITESAMPLE] = "writesample", \
2373 [OBD_HEAT_READBYTE] = "readbyte", \
2374 [OBD_HEAT_WRITEBYTE] = "writebyte", \
2389 static inline const char *pcc_type2string(enum lu_pcc_type type)
2394 case LU_PCC_READWRITE:
2401 struct lu_pcc_attach {
2402 __u32 pcca_type; /* PCC type */
2403 __u32 pcca_id; /* archive ID for readwrite, group ID for readonly */
2406 enum lu_pcc_detach_opts {
2407 PCC_DETACH_OPT_NONE = 0, /* Detach only, keep the PCC copy */
2408 PCC_DETACH_OPT_UNCACHE, /* Remove the cached file after detach */
2411 struct lu_pcc_detach_fid {
2412 /* fid of the file to detach */
2413 struct lu_fid pccd_fid;
2417 struct lu_pcc_detach {
2421 enum lu_pcc_state_flags {
2422 PCC_STATE_FL_NONE = 0x0,
2423 /* The inode attr is cached locally */
2424 PCC_STATE_FL_ATTR_VALID = 0x01,
2425 /* The file is being attached into PCC */
2426 PCC_STATE_FL_ATTACHING = 0x02,
2427 /* Allow to auto attach at open */
2428 PCC_STATE_FL_OPEN_ATTACH = 0x04,
2431 struct lu_pcc_state {
2432 __u32 pccs_type; /* enum lu_pcc_type */
2433 __u32 pccs_open_count;
2434 __u32 pccs_flags; /* enum lu_pcc_state_flags */
2436 char pccs_path[PATH_MAX];
2441 /* make header's size equal lu_fid */
2444 struct lu_fid fa_fids[0];
2446 #define OBD_MAX_FIDS_IN_ARRAY 4096
2448 #if defined(__cplusplus)
2452 /** @} lustreuser */
2454 #endif /* _LUSTRE_USER_H */