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, 2016, 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/types.h>
48 # include <linux/fs.h>
49 # include <linux/quota.h>
50 # include <linux/string.h> /* snprintf() */
51 # include <linux/version.h>
52 # include <uapi/linux/lustre/lustre_fiemap.h>
53 #else /* !__KERNEL__ */
54 # define NEED_QUOTA_DEFS
57 # include <stdio.h> /* snprintf() */
59 # include <sys/quota.h>
60 # include <sys/stat.h>
61 # include <linux/lustre/lustre_fiemap.h>
62 #endif /* __KERNEL__ */
64 #if defined(__cplusplus)
69 * This is a temporary solution of adding quota type.
70 * Should be removed as soon as system header is updated.
73 #define LL_MAXQUOTAS 3
75 #define INITQFNAMES { \
76 "user", /* USRQUOTA */ \
77 "group", /* GRPQUOTA */ \
78 "project", /* PRJQUOTA */ \
83 #if defined(__x86_64__) || defined(__ia64__) || defined(__ppc64__) || \
84 defined(__craynv) || defined(__mips64__) || defined(__powerpc64__) || \
86 typedef struct stat lstat_t;
87 # define lstat_f lstat
88 # define fstat_f fstat
89 # define fstatat_f fstatat
90 # define HAVE_LOV_USER_MDS_DATA
91 #elif defined(__USE_LARGEFILE64) || defined(__KERNEL__)
92 typedef struct stat64 lstat_t;
93 # define lstat_f lstat64
94 # define fstat_f fstat64
95 # define fstatat_f fstatat64
96 # define HAVE_LOV_USER_MDS_DATA
99 #define LUSTRE_EOF 0xffffffffffffffffULL
102 #define LL_SUPER_MAGIC 0x0BD00BD0
104 #ifndef FSFILT_IOC_GETFLAGS
105 #define FSFILT_IOC_GETFLAGS _IOR('f', 1, long)
106 #define FSFILT_IOC_SETFLAGS _IOW('f', 2, long)
107 #define FSFILT_IOC_GETVERSION _IOR('f', 3, long)
108 #define FSFILT_IOC_SETVERSION _IOW('f', 4, long)
109 #define FSFILT_IOC_GETVERSION_OLD _IOR('v', 1, long)
110 #define FSFILT_IOC_SETVERSION_OLD _IOW('v', 2, long)
113 /* FIEMAP flags supported by Lustre */
114 #define LUSTRE_FIEMAP_FLAGS_COMPAT (FIEMAP_FLAG_SYNC | FIEMAP_FLAG_DEVICE_ORDER)
116 enum obd_statfs_state {
117 OS_STATE_DEGRADED = 0x00000001, /**< RAID degraded/rebuilding */
118 OS_STATE_READONLY = 0x00000002, /**< filesystem is read-only */
119 OS_STATE_ENOSPC = 0x00000020, /**< not enough free space */
120 OS_STATE_ENOINO = 0x00000040, /**< not enough inodes */
123 /** filesystem statistics/attributes for target device */
125 __u64 os_type; /* EXT4_SUPER_MAGIC, UBERBLOCK_MAGIC */
126 __u64 os_blocks; /* total size in #os_bsize blocks */
127 __u64 os_bfree; /* number of unused blocks */
128 __u64 os_bavail; /* blocks available for allocation */
129 __u64 os_files; /* total number of objects */
130 __u64 os_ffree; /* # objects that could be created */
131 __u8 os_fsid[40]; /* identifier for filesystem */
132 __u32 os_bsize; /* block size in bytes for os_blocks */
133 __u32 os_namelen; /* maximum length of filename in bytes*/
134 __u64 os_maxbytes; /* maximum object size in bytes */
135 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
136 __u32 os_fprecreated; /* objs available now to the caller */
137 /* used in QoS code to find preferred
139 __u32 os_spare2; /* Unused padding fields. Remember */
140 __u32 os_spare3; /* to fix lustre_swab_obd_statfs() */
152 * FID is a cluster-wide unique identifier of a file or an object (stripe).
153 * FIDs are never reused.
157 * FID sequence. Sequence is a unit of migration: all files (objects)
158 * with FIDs from a given sequence are stored on the same server.
159 * Lustre should support 2^64 objects, so even if each sequence
160 * has only a single object we can still enumerate 2^64 objects.
163 /* FID number within sequence. */
166 * FID version, used to distinguish different versions (in the sense
167 * of snapshots, etc.) of the same file system object. Not currently
173 static inline bool fid_is_zero(const struct lu_fid *fid)
175 return fid->f_seq == 0 && fid->f_oid == 0;
178 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
179 * MDT-object's FID::f_ver, instead it is the OST-object index in its
180 * parent MDT-object's layout EA. */
181 #define f_stripe_idx f_ver
184 __u32 ol_stripe_size;
185 __u32 ol_stripe_count;
189 } __attribute__((packed));
191 /* keep this one for compatibility */
192 struct filter_fid_old {
193 struct lu_fid ff_parent;
199 struct lu_fid ff_parent;
200 struct ost_layout ff_layout;
201 __u32 ff_layout_version;
202 __u32 ff_range; /* range of layout version that
203 * write are allowed */
204 } __attribute__((packed));
206 /* Userspace should treat lu_fid as opaque, and only use the following methods
207 * to print or parse them. Other functions (e.g. compare, swab) could be moved
208 * here from lustre_idl.h if needed. */
212 LMAC_HSM = 0x00000001,
213 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
214 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
215 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
216 * under /O/<seq>/d<x>. */
217 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
218 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
222 * Masks for all features that should be supported by a Lustre version to
223 * access a specific file.
224 * This information is stored in lustre_mdt_attrs::lma_incompat.
227 LMAI_RELEASED = 0x00000001, /* file is released */
228 LMAI_AGENT = 0x00000002, /* agent inode */
229 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
230 is on the remote MDT */
231 LMAI_STRIPED = 0x00000008, /* striped directory inode */
232 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
233 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
234 LMAI_STRIPED | LMAI_ORPHAN)
239 * Following struct for object attributes, that will be kept inode's EA.
240 * Introduced in 2.0 release (please see b15993, for details)
241 * Added to all objects since Lustre 2.4 as contains self FID
243 struct lustre_mdt_attrs {
245 * Bitfield for supported data in this structure. From enum lma_compat.
246 * lma_self_fid and lma_flags are always available.
250 * Per-file incompat feature list. Lustre version should support all
251 * flags set in this field. The supported feature mask is available in
255 /** FID of this inode */
256 struct lu_fid lma_self_fid;
259 struct lustre_ost_attrs {
260 /* Use lustre_mdt_attrs directly for now, need a common header
261 * structure if want to change lustre_mdt_attrs in future. */
262 struct lustre_mdt_attrs loa_lma;
264 /* Below five elements are for OST-object's PFID EA, the
265 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
266 * and the stripe_index (low 16 bits), the size should not exceed
267 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
268 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
269 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
270 * loa_comp_* elements are valid. */
271 struct lu_fid loa_parent_fid;
272 __u32 loa_stripe_size;
274 __u64 loa_comp_start;
279 * Prior to 2.4, the LMA structure also included SOM attributes which has since
280 * been moved to a dedicated xattr
281 * lma_flags was also removed because of lma_compat/incompat fields.
283 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
286 LSOM_FL_VALID = 1 << 0,
289 struct lustre_som_attrs {
291 __u16 lsa_reserved[3];
297 * OST object IDentifier.
305 struct lu_fid oi_fid;
309 #define DOSTID "%#llx:%llu"
310 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
311 ((unsigned long long)ostid_id(oi))
313 struct ll_futimes_3 {
315 __u64 lfu_atime_nsec;
317 __u64 lfu_mtime_nsec;
319 __u64 lfu_ctime_nsec;
323 * Maximum number of mirrors currently implemented.
325 #define LUSTRE_MIRROR_COUNT_MAX 16
327 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
329 LL_LEASE_RDLCK = 0x01,
330 LL_LEASE_WRLCK = 0x02,
331 LL_LEASE_UNLCK = 0x04,
334 enum ll_lease_flags {
335 LL_LEASE_RESYNC = 0x1,
336 LL_LEASE_RESYNC_DONE = 0x2,
339 #define IOC_IDS_MAX 4096
340 struct ll_ioc_lease {
348 * The ioctl naming rules:
349 * LL_* - works on the currently opened filehandle instead of parent dir
350 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
351 * *_MDC_* - gets/sets data related to MDC
352 * *_LOV_* - gets/sets data related to OSC/LOV
353 * *FILE* - called on parent dir and passes in a filename
354 * *STRIPE* - set/get lov_user_md
355 * *INFO - set/get lov_user_mds_data
357 /* lustre_ioctl.h 101-150 */
358 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
359 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
360 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
361 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
362 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
363 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
364 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
365 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
366 /* LL_IOC_RECREATE_OBJ 157 obsolete */
367 /* LL_IOC_RECREATE_FID 157 obsolete */
368 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
369 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
370 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
371 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
372 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
373 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
374 /* IOC_LOV_GETINFO 165 obsolete */
375 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
376 /* LL_IOC_RMTACL 167 obsolete */
377 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
378 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
379 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
380 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
381 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
382 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
383 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
384 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
385 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
386 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
387 /* lustre_ioctl.h 177-210 */
388 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
389 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
390 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
391 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
392 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
393 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
394 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
395 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
396 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
397 struct lustre_swap_layouts)
398 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
399 struct hsm_current_action)
400 /* lustre_ioctl.h 221-232 */
401 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
402 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
403 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
404 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
405 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
406 #define LL_IOC_GET_LEASE _IO('f', 244)
407 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
408 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
409 #define LL_IOC_MIGRATE _IOR('f', 247, int)
410 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
411 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
412 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
414 #ifndef FS_IOC_FSGETXATTR
416 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
419 __u32 fsx_xflags; /* xflags field value (get/set) */
420 __u32 fsx_extsize; /* extsize field value (get/set)*/
421 __u32 fsx_nextents; /* nextents field value (get) */
422 __u32 fsx_projid; /* project identifier (get/set) */
423 unsigned char fsx_pad[12];
425 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
426 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
428 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
429 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
430 #define LL_PROJINHERIT_FL 0x20000000
433 #define LL_STATFS_LMV 1
434 #define LL_STATFS_LOV 2
435 #define LL_STATFS_NODELAY 4
437 #define IOC_MDC_TYPE 'i'
438 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
439 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
440 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
441 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
443 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
445 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
446 * files, but are unlikely to be used in practice and are not harmful if
447 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
448 * devices and are safe for use on new files. See LU-4209. */
449 /* To be compatible with old statically linked binary we keep the check for
450 * the older 0100000000 flag. This is already removed upstream. LU-812. */
451 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
452 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
453 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
454 O_LOV_DELAY_CREATE_MASK)
456 #define LL_FILE_IGNORE_LOCK 0x00000001
457 #define LL_FILE_GROUP_LOCKED 0x00000002
458 #define LL_FILE_READAHEA 0x00000004
459 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
460 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
462 #define LOV_USER_MAGIC_V1 0x0BD10BD0
463 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
464 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
465 #define LOV_USER_MAGIC_V3 0x0BD30BD0
466 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
467 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
468 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
470 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
471 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
472 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
474 #define LOV_PATTERN_NONE 0x000
475 #define LOV_PATTERN_RAID0 0x001
476 #define LOV_PATTERN_RAID1 0x002
477 #define LOV_PATTERN_MDT 0x100
478 #define LOV_PATTERN_CMOBD 0x200
480 #define LOV_PATTERN_F_MASK 0xffff0000
481 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
482 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
483 #define LOV_PATTERN_DEFAULT 0xffffffff
485 static inline bool lov_pattern_supported(__u32 pattern)
487 return pattern == LOV_PATTERN_RAID0 ||
488 pattern == LOV_PATTERN_MDT ||
489 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_F_RELEASED);
492 #define LOV_MAXPOOLNAME 15
493 #define LOV_POOLNAMEF "%.15s"
495 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
496 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
497 #define LOV_MAX_STRIPE_COUNT_OLD 160
498 /* This calculation is crafted so that input of 4096 will result in 160
499 * which in turn is equal to old maximal stripe count.
500 * XXX: In fact this is too simpified for now, what it also need is to get
501 * ea_type argument to clearly know how much space each stripe consumes.
503 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
504 * allocation that is sufficient for the current generation of systems.
506 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
507 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
508 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
509 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
511 #define XATTR_LUSTRE_PREFIX "lustre."
512 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
514 #define lov_user_ost_data lov_user_ost_data_v1
515 struct lov_user_ost_data_v1 { /* per-stripe data structure */
516 struct ost_id l_ost_oi; /* OST object ID */
517 __u32 l_ost_gen; /* generation of this OST index */
518 __u32 l_ost_idx; /* OST index in LOV */
519 } __attribute__((packed));
521 #define lov_user_md lov_user_md_v1
522 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
523 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
524 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
525 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
526 __u32 lmm_stripe_size; /* size of stripe in bytes */
527 __u16 lmm_stripe_count; /* num stripes in use for this object */
529 __u16 lmm_stripe_offset; /* starting stripe offset in
530 * lmm_objects, use when writing */
531 __u16 lmm_layout_gen; /* layout generation number
532 * used when reading */
534 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
535 } __attribute__((packed, __may_alias__));
537 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
538 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
539 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
540 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
541 __u32 lmm_stripe_size; /* size of stripe in bytes */
542 __u16 lmm_stripe_count; /* num stripes in use for this object */
544 __u16 lmm_stripe_offset; /* starting stripe offset in
545 * lmm_objects, use when writing */
546 __u16 lmm_layout_gen; /* layout generation number
547 * used when reading */
549 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
550 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
551 } __attribute__((packed));
558 #define DEXT "[%#llx, %#llx)"
559 #define PEXT(ext) (ext)->e_start, (ext)->e_end
561 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
562 struct lu_extent *e2)
564 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
567 static inline bool lu_extent_is_whole(struct lu_extent *e)
569 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
572 enum lov_comp_md_entry_flags {
573 LCME_FL_PRIMARY = 0x00000001, /* Not used */
574 LCME_FL_STALE = 0x00000002, /* Not used */
575 LCME_FL_OFFLINE = 0x00000004, /* Not used */
576 LCME_FL_PREFERRED = 0x00000008, /* Not used */
577 LCME_FL_INIT = 0x00000010, /* instantiated */
578 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
579 won't be stored on disk */
582 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT)
584 /* the highest bit in obdo::o_layout_version is used to mark if the file is
586 #define LU_LAYOUT_RESYNC LCME_FL_NEG
588 /* lcme_id can be specified as certain flags, and the the first
589 * bit of lcme_id is used to indicate that the ID is representing
590 * certain LCME_FL_* but not a real ID. Which implies we can have
591 * at most 31 flags (see LCME_FL_XXX). */
594 LCME_ID_MAX = 0x7FFFFFFF,
595 LCME_ID_ALL = 0xFFFFFFFF,
596 LCME_ID_NOT_ID = LCME_FL_NEG
599 #define LCME_ID_MASK LCME_ID_MAX
601 struct lov_comp_md_entry_v1 {
602 __u32 lcme_id; /* unique id of component */
603 __u32 lcme_flags; /* LCME_FL_XXX */
604 struct lu_extent lcme_extent; /* file extent for component */
605 __u32 lcme_offset; /* offset of component blob,
606 start from lov_comp_md_v1 */
607 __u32 lcme_size; /* size of component blob */
608 __u64 lcme_padding[2];
609 } __attribute__((packed));
611 #define SEQ_ID_MAX 0x0000FFFF
612 #define SEQ_ID_MASK SEQ_ID_MAX
613 /* bit 30:16 of lcme_id is used to store mirror id */
614 #define MIRROR_ID_MASK 0x7FFF0000
615 #define MIRROR_ID_SHIFT 16
617 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
619 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
622 static inline __u16 mirror_id_of(__u32 id)
624 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
628 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
630 enum lov_comp_md_flags {
631 /* the least 2 bits are used by FLR to record file state */
634 LCM_FL_WRITE_PENDING = 2,
635 LCM_FL_SYNC_PENDING = 3,
636 LCM_FL_FLR_MASK = 0x3,
639 struct lov_comp_md_v1 {
640 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
641 __u32 lcm_size; /* overall size including this struct */
642 __u32 lcm_layout_gen;
644 __u16 lcm_entry_count;
645 /* lcm_mirror_count stores the number of actual mirrors minus 1,
646 * so that non-flr files will have value 0 meaning 1 mirror. */
647 __u16 lcm_mirror_count;
648 __u16 lcm_padding1[3];
650 struct lov_comp_md_entry_v1 lcm_entries[0];
651 } __attribute__((packed));
654 * Maximum number of mirrors Lustre can support.
656 #define LUSTRE_MIRROR_COUNT_MAX 16
658 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
660 if (stripes == (__u16)-1)
663 if (lmm_magic == LOV_USER_MAGIC_V1)
664 return sizeof(struct lov_user_md_v1) +
665 stripes * sizeof(struct lov_user_ost_data_v1);
666 return sizeof(struct lov_user_md_v3) +
667 stripes * sizeof(struct lov_user_ost_data_v1);
670 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
671 * use this. It is unsafe to #define those values in this header as it
672 * is possible the application has already #included <sys/stat.h>. */
673 #ifdef HAVE_LOV_USER_MDS_DATA
674 #define lov_user_mds_data lov_user_mds_data_v1
675 struct lov_user_mds_data_v1 {
676 lstat_t lmd_st; /* MDS stat struct */
677 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
678 } __attribute__((packed));
680 struct lov_user_mds_data_v3 {
681 lstat_t lmd_st; /* MDS stat struct */
682 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
683 } __attribute__((packed));
686 struct lmv_user_mds_data {
687 struct lu_fid lum_fid;
693 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
694 LMV_HASH_TYPE_ALL_CHARS = 1,
695 LMV_HASH_TYPE_FNV_1A_64 = 2,
699 #define LMV_HASH_NAME_ALL_CHARS "all_char"
700 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
702 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
704 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
705 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
706 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
707 #define lmv_user_md lmv_user_md_v1
708 struct lmv_user_md_v1 {
709 __u32 lum_magic; /* must be the first field */
710 __u32 lum_stripe_count; /* dirstripe count */
711 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
712 __u32 lum_hash_type; /* Dir stripe policy */
713 __u32 lum_type; /* LMV type: default or normal */
717 char lum_pool_name[LOV_MAXPOOLNAME + 1];
718 struct lmv_user_mds_data lum_objects[0];
719 } __attribute__((packed));
721 static inline int lmv_user_md_size(int stripes, int lmm_magic)
723 return sizeof(struct lmv_user_md) +
724 stripes * sizeof(struct lmv_user_mds_data);
727 struct ll_recreate_obj {
733 __u64 id; /* holds object id */
734 __u32 generation; /* holds object generation */
735 __u32 f_type; /* holds object type or stripe idx when passing it to
736 * OST for saving into EA. */
744 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
745 const struct obd_uuid *u2)
747 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
750 static inline int obd_uuid_empty(struct obd_uuid *uuid)
752 return uuid->uuid[0] == '\0';
755 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
757 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
758 uuid->uuid[sizeof(*uuid) - 1] = '\0';
761 /* For printf's only, make sure uuid is terminated */
762 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
767 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
768 /* Obviously not safe, but for printfs, no real harm done...
769 we're always null-terminated, even in a race. */
770 static char temp[sizeof(*uuid)];
771 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
772 temp[sizeof(*uuid) - 1] = '\0';
775 return (char *)(uuid->uuid);
778 #define LUSTRE_MAXFSNAME 8
780 /* Extract fsname from uuid (or target name) of a target
781 e.g. (myfs-OST0007_UUID -> myfs)
782 see also deuuidify. */
783 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
787 strncpy(buf, uuid, buflen - 1);
788 buf[buflen - 1] = '\0';
789 p = strrchr(buf, '-');
794 /* printf display format for Lustre FIDs
795 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
796 #define FID_NOBRACE_LEN 40
797 #define FID_LEN (FID_NOBRACE_LEN + 2)
798 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
799 #define DFID "["DFID_NOBRACE"]"
800 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
802 /* scanf input parse format for fids in DFID_NOBRACE format
803 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
804 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
805 #define SFID "0x%llx:0x%x:0x%x"
806 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
808 /********* Quotas **********/
810 #define LUSTRE_QUOTABLOCK_BITS 10
811 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
813 static inline __u64 lustre_stoqb(size_t space)
815 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
818 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
819 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
820 #define Q_GETOINFO 0x800102 /* get obd quota info */
821 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
822 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
824 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
825 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
826 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
827 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
828 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
829 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
830 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
831 /* lustre-specific control commands */
832 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
833 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
835 #define ALLQUOTA 255 /* set all quota */
836 static inline char *qtype_name(int qtype)
849 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
852 #define N_PERMS_MAX 64
854 struct perm_downcall_data {
860 struct identity_downcall_data {
867 struct perm_downcall_data idd_perms[N_PERMS_MAX];
871 #ifdef NEED_QUOTA_DEFS
873 #define QIF_BLIMITS 1
875 #define QIF_ILIMITS 4
879 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
880 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
881 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
882 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
885 #endif /* !__KERNEL__ */
887 /* lustre volatile file support
888 * file name header: .^L^S^T^R:volatile"
890 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
891 #define LUSTRE_VOLATILE_HDR_LEN 14
893 typedef enum lustre_quota_version {
895 } lustre_quota_version_t;
897 /* XXX: same as if_dqinfo struct in kernel */
905 /* XXX: same as if_dqblk struct in kernel, plus one padding */
907 __u64 dqb_bhardlimit;
908 __u64 dqb_bsoftlimit;
910 __u64 dqb_ihardlimit;
911 __u64 dqb_isoftlimit;
933 struct obd_dqinfo qc_dqinfo;
934 struct obd_dqblk qc_dqblk;
936 struct obd_uuid obd_uuid;
939 /* swap layout flags */
940 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
941 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
942 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
943 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
944 #define SWAP_LAYOUTS_CLOSE (1 << 4)
945 #define MERGE_LAYOUTS_CLOSE (1 << 5)
946 #define INTENT_LAYOUTS_CLOSE (SWAP_LAYOUTS_CLOSE | MERGE_LAYOUTS_CLOSE)
948 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
949 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
950 struct lustre_swap_layouts {
959 /********* Changelogs **********/
960 /** Changelog record types */
961 enum changelog_rec_type {
963 CL_CREATE = 1, /* namespace */
964 CL_MKDIR = 2, /* namespace */
965 CL_HARDLINK = 3, /* namespace */
966 CL_SOFTLINK = 4, /* namespace */
967 CL_MKNOD = 5, /* namespace */
968 CL_UNLINK = 6, /* namespace */
969 CL_RMDIR = 7, /* namespace */
970 CL_RENAME = 8, /* namespace */
971 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
972 CL_OPEN = 10, /* not currently used */
973 CL_CLOSE = 11, /* may be written to log only with mtime change */
974 CL_LAYOUT = 12, /* file layout/striping modified */
978 CL_HSM = 16, /* HSM specific events, see flags */
979 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
983 CL_FLRW = 21, /* FLR: file was firstly written */
984 CL_RESYNC = 22, /* FLR: file was resync-ed */
988 static inline const char *changelog_type2str(int type) {
989 static const char *changelog_str[] = {
990 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
991 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
992 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
996 if (type >= 0 && type < CL_LAST)
997 return changelog_str[type];
1001 /* per-record flags */
1002 #define CLF_FLAGSHIFT 12
1003 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
1004 #define CLF_VERMASK (~CLF_FLAGMASK)
1005 enum changelog_rec_flags {
1006 CLF_VERSION = 0x1000,
1007 CLF_RENAME = 0x2000,
1009 CLF_EXTRA_FLAGS = 0x8000,
1010 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS
1014 /* Anything under the flagmask may be per-type (if desired) */
1015 /* Flags for unlink */
1016 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1017 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1018 /* HSM cleaning needed */
1019 /* Flags for rename */
1020 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1022 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1023 * has an archive in backend */
1026 /* 12b used (from high weight to low weight):
1031 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1032 #define CLF_HSM_ERR_H 6
1033 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1034 #define CLF_HSM_EVENT_H 9
1035 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1036 #define CLF_HSM_FLAG_H 11
1037 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1038 #define CLF_HSM_SPARE_H 15
1039 #define CLF_HSM_LAST 15
1041 /* Remove bits higher than _h, then extract the value
1042 * between _h and _l by shifting lower weigth to bit 0. */
1043 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1044 >> (CLF_HSM_LAST - _h + _l))
1046 #define CLF_HSM_SUCCESS 0x00
1047 #define CLF_HSM_MAXERROR 0x7E
1048 #define CLF_HSM_ERROVERFLOW 0x7F
1050 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1052 /* 3 bits field => 8 values allowed */
1064 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1066 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1070 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
1072 *flags |= (he << CLF_HSM_EVENT_L);
1075 static inline __u16 hsm_get_cl_flags(int flags)
1077 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1080 static inline void hsm_set_cl_flags(int *flags, int bits)
1082 *flags |= (bits << CLF_HSM_FLAG_L);
1085 static inline int hsm_get_cl_error(int flags)
1087 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1090 static inline void hsm_set_cl_error(int *flags, int error)
1092 *flags |= (error << CLF_HSM_ERR_L);
1095 enum changelog_rec_extra_flags {
1097 CLFE_UIDGID = 0x0001,
1098 CLFE_SUPPORTED = CLFE_UIDGID
1101 enum changelog_send_flag {
1102 /* Not yet implemented */
1103 CHANGELOG_FLAG_FOLLOW = 0x01,
1104 /* Blocking IO makes sense in case of slow user parsing of the records,
1105 * but it also prevents us from cleaning up if the records are not
1107 CHANGELOG_FLAG_BLOCK = 0x02,
1108 /* Pack jobid into the changelog records if available. */
1109 CHANGELOG_FLAG_JOBID = 0x04,
1110 /* Pack additional flag bits into the changelog record */
1111 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1114 enum changelog_send_extra_flag {
1115 /* Pack uid/gid into the changelog record */
1116 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1119 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
1120 changelog_rec_offset(CLF_SUPPORTED, \
1123 /* 31 usable bytes string + null terminator. */
1124 #define LUSTRE_JOBID_SIZE 32
1126 /* This is the minimal changelog record. It can contain extensions
1127 * such as rename fields or process jobid. Its exact content is described
1128 * by the cr_flags and cr_extra_flags.
1130 * Extensions are packed in the same order as their corresponding flags,
1131 * then in the same order as their corresponding extra flags.
1133 struct changelog_rec {
1135 __u16 cr_flags; /**< \a changelog_rec_flags */
1136 __u32 cr_type; /**< \a changelog_rec_type */
1137 __u64 cr_index; /**< changelog record number */
1138 __u64 cr_prev; /**< last index for this target fid */
1141 struct lu_fid cr_tfid; /**< target fid */
1142 __u32 cr_markerflags; /**< CL_MARK flags */
1144 struct lu_fid cr_pfid; /**< parent fid */
1147 /* Changelog extension for RENAME. */
1148 struct changelog_ext_rename {
1149 struct lu_fid cr_sfid; /**< source fid, or zero */
1150 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1153 /* Changelog extension to include JOBID. */
1154 struct changelog_ext_jobid {
1155 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1158 /* Changelog extension to include additional flags. */
1159 struct changelog_ext_extra_flags {
1160 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1163 /* Changelog extra extension to include UID/GID. */
1164 struct changelog_ext_uidgid {
1169 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1170 const struct changelog_rec *rec);
1172 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1173 enum changelog_rec_extra_flags cref)
1175 size_t size = sizeof(struct changelog_rec);
1177 if (crf & CLF_RENAME)
1178 size += sizeof(struct changelog_ext_rename);
1180 if (crf & CLF_JOBID)
1181 size += sizeof(struct changelog_ext_jobid);
1183 if (crf & CLF_EXTRA_FLAGS) {
1184 size += sizeof(struct changelog_ext_extra_flags);
1185 if (cref & CLFE_UIDGID)
1186 size += sizeof(struct changelog_ext_uidgid);
1192 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1194 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1196 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1197 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1199 return changelog_rec_offset(rec->cr_flags, cref);
1202 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1204 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1208 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1210 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1212 return (struct changelog_ext_rename *)((char *)rec +
1213 changelog_rec_offset(crf,
1217 /* The jobid follows the rename extension, if present */
1219 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1221 enum changelog_rec_flags crf = rec->cr_flags &
1222 (CLF_VERSION | CLF_RENAME);
1224 return (struct changelog_ext_jobid *)((char *)rec +
1225 changelog_rec_offset(crf,
1229 /* The additional flags follow the rename and jobid extensions, if present */
1231 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1232 const struct changelog_rec *rec)
1234 enum changelog_rec_flags crf = rec->cr_flags &
1235 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1237 return (struct changelog_ext_extra_flags *)((char *)rec +
1238 changelog_rec_offset(crf,
1242 /* The uid/gid is the first extra extension */
1244 struct changelog_ext_uidgid *changelog_rec_uidgid(
1245 const struct changelog_rec *rec)
1247 enum changelog_rec_flags crf = rec->cr_flags &
1248 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1250 return (struct changelog_ext_uidgid *)((char *)rec +
1251 changelog_rec_offset(crf,
1255 /* The name follows the rename, jobid and extra flags extns, if present */
1256 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1258 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1260 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1261 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1263 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1264 cref & CLFE_SUPPORTED);
1267 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1269 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1272 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1274 char *cr_name = changelog_rec_name(rec);
1276 return cr_name + strlen(cr_name) + 1;
1280 * Remap a record to the desired format as specified by the crf flags.
1281 * The record must be big enough to contain the final remapped version.
1282 * Superfluous extension fields are removed and missing ones are added
1283 * and zeroed. The flags of the record are updated accordingly.
1285 * The jobid and rename extensions can be added to a record, to match the
1286 * format an application expects, typically. In this case, the newly added
1287 * fields will be zeroed.
1288 * The Jobid field can be removed, to guarantee compatibility with older
1289 * clients that don't expect this field in the records they process.
1291 * The following assumptions are being made:
1292 * - CLF_RENAME will not be removed
1293 * - CLF_JOBID will not be added without CLF_RENAME being added too
1294 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1296 * @param[in,out] rec The record to remap.
1297 * @param[in] crf_wanted Flags describing the desired extensions.
1298 * @param[in] cref_want Flags describing the desired extra extensions.
1300 static inline void changelog_remap_rec(struct changelog_rec *rec,
1301 enum changelog_rec_flags crf_wanted,
1302 enum changelog_rec_extra_flags cref_want)
1304 char *uidgid_mov = NULL;
1308 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1310 crf_wanted &= CLF_SUPPORTED;
1311 cref_want &= CLFE_SUPPORTED;
1313 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1314 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1315 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1316 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1322 /* First move the variable-length name field */
1323 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1324 changelog_rec_name(rec), rec->cr_namelen);
1326 /* Locations of extensions in the remapped record */
1327 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1328 uidgid_mov = (char *)rec +
1329 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1331 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1334 ef_mov = (char *)rec +
1335 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1337 jid_mov = (char *)rec +
1338 changelog_rec_offset(crf_wanted &
1339 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1341 rnm_mov = (char *)rec +
1342 changelog_rec_offset(crf_wanted &
1348 /* Move the extension fields to the desired positions */
1349 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1350 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1351 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1352 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1353 sizeof(struct changelog_ext_uidgid));
1355 memmove(ef_mov, changelog_rec_extra_flags(rec),
1356 sizeof(struct changelog_ext_extra_flags));
1359 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1360 memmove(jid_mov, changelog_rec_jobid(rec),
1361 sizeof(struct changelog_ext_jobid));
1363 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1364 memmove(rnm_mov, changelog_rec_rename(rec),
1365 sizeof(struct changelog_ext_rename));
1367 /* Clear newly added fields */
1368 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1369 !(cref & CLFE_UIDGID))
1370 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1372 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1373 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1374 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1376 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1377 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1379 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1380 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1382 /* Update the record's flags accordingly */
1383 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1384 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1385 changelog_rec_extra_flags(rec)->cr_extra_flags =
1386 changelog_rec_extra_flags(rec)->cr_extra_flags |
1390 enum changelog_message_type {
1391 CL_RECORD = 10, /* message is a changelog_rec */
1392 CL_EOF = 11, /* at end of current changelog */
1395 /********* Misc **********/
1397 struct ioc_data_version {
1399 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1400 __u32 idv_flags; /* enum ioc_data_version_flags */
1403 enum ioc_data_version_flags {
1404 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1405 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1409 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1412 #define dot_lustre_name ".lustre"
1415 /********* HSM **********/
1417 /** HSM per-file state
1418 * See HSM_FLAGS below.
1421 HS_NONE = 0x00000000,
1422 HS_EXISTS = 0x00000001,
1423 HS_DIRTY = 0x00000002,
1424 HS_RELEASED = 0x00000004,
1425 HS_ARCHIVED = 0x00000008,
1426 HS_NORELEASE = 0x00000010,
1427 HS_NOARCHIVE = 0x00000020,
1428 HS_LOST = 0x00000040,
1431 /* HSM user-setable flags. */
1432 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1434 /* Other HSM flags. */
1435 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1438 * All HSM-related possible flags that could be applied to a file.
1439 * This should be kept in sync with hsm_states.
1441 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1444 * HSMÂ request progress state
1446 enum hsm_progress_states {
1453 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1456 case HPS_WAITING: return "waiting";
1457 case HPS_RUNNING: return "running";
1458 case HPS_DONE: return "done";
1459 default: return "unknown";
1466 } __attribute__((packed));
1469 * Current HSM states of a Lustre file.
1471 * This structure purpose is to be sent to user-space mainly. It describes the
1472 * current HSM flags and in-progress action.
1474 struct hsm_user_state {
1475 /** Current HSM states, from enum hsm_states. */
1477 __u32 hus_archive_id;
1478 /** The current undergoing action, if there is one */
1479 __u32 hus_in_progress_state;
1480 __u32 hus_in_progress_action;
1481 struct hsm_extent hus_in_progress_location;
1482 char hus_extended_info[];
1485 struct hsm_state_set_ioc {
1486 struct lu_fid hssi_fid;
1488 __u64 hssi_clearmask;
1492 * This structure describes the current in-progress action for a file.
1493 * it is retuned to user space and send over the wire
1495 struct hsm_current_action {
1496 /** The current undergoing action, if there is one */
1497 /* state is one of hsm_progress_states */
1499 /* action is one of hsm_user_action */
1501 struct hsm_extent hca_location;
1504 /***** HSM user requests ******/
1505 /* User-generated (lfs/ioctl) request types */
1506 enum hsm_user_action {
1507 HUA_NONE = 1, /* no action (noop) */
1508 HUA_ARCHIVE = 10, /* copy to hsm */
1509 HUA_RESTORE = 11, /* prestage */
1510 HUA_RELEASE = 12, /* drop ost objects */
1511 HUA_REMOVE = 13, /* remove from archive */
1512 HUA_CANCEL = 14 /* cancel a request */
1515 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1518 case HUA_NONE: return "NOOP";
1519 case HUA_ARCHIVE: return "ARCHIVE";
1520 case HUA_RESTORE: return "RESTORE";
1521 case HUA_RELEASE: return "RELEASE";
1522 case HUA_REMOVE: return "REMOVE";
1523 case HUA_CANCEL: return "CANCEL";
1524 default: return "UNKNOWN";
1529 * List of hr_flags (bit field)
1531 #define HSM_FORCE_ACTION 0x0001
1532 /* used by CT, cannot be set by user */
1533 #define HSM_GHOST_COPY 0x0002
1536 * Contains all the fixed part of struct hsm_user_request.
1539 struct hsm_request {
1540 __u32 hr_action; /* enum hsm_user_action */
1541 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1542 __u64 hr_flags; /* request flags */
1543 __u32 hr_itemcount; /* item count in hur_user_item vector */
1547 struct hsm_user_item {
1548 struct lu_fid hui_fid;
1549 struct hsm_extent hui_extent;
1550 } __attribute__((packed));
1552 struct hsm_user_request {
1553 struct hsm_request hur_request;
1554 struct hsm_user_item hur_user_item[0];
1555 /* extra data blob at end of struct (after all
1556 * hur_user_items), only use helpers to access it
1558 } __attribute__((packed));
1560 /** Return pointer to data field in a hsm user request */
1561 static inline void *hur_data(struct hsm_user_request *hur)
1563 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1567 * Compute the current length of the provided hsm_user_request. This returns -1
1568 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1570 * return -1 on bounds check error.
1572 static inline ssize_t hur_len(struct hsm_user_request *hur)
1576 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1577 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1578 (__u64)hur->hur_request.hr_itemcount *
1579 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1581 if (size != (ssize_t)size)
1587 /****** HSM RPCs to copytool *****/
1588 /* Message types the copytool may receive */
1589 enum hsm_message_type {
1590 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1593 /* Actions the copytool may be instructed to take for a given action_item */
1594 enum hsm_copytool_action {
1595 HSMA_NONE = 10, /* no action */
1596 HSMA_ARCHIVE = 20, /* arbitrary offset */
1602 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1605 case HSMA_NONE: return "NOOP";
1606 case HSMA_ARCHIVE: return "ARCHIVE";
1607 case HSMA_RESTORE: return "RESTORE";
1608 case HSMA_REMOVE: return "REMOVE";
1609 case HSMA_CANCEL: return "CANCEL";
1610 default: return "UNKNOWN";
1614 /* Copytool item action description */
1615 struct hsm_action_item {
1616 __u32 hai_len; /* valid size of this struct */
1617 __u32 hai_action; /* hsm_copytool_action, but use known size */
1618 struct lu_fid hai_fid; /* Lustre FID to operate on */
1619 struct lu_fid hai_dfid; /* fid used for data access */
1620 struct hsm_extent hai_extent; /* byte range to operate on */
1621 __u64 hai_cookie; /* action cookie from coordinator */
1622 __u64 hai_gid; /* grouplock id */
1623 char hai_data[0]; /* variable length */
1624 } __attribute__((packed));
1627 * helper function which print in hexa the first bytes of
1630 * \param hai [IN] record to print
1631 * \param buffer [IN,OUT] buffer to write the hex string to
1632 * \param len [IN] max buffer length
1636 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1637 char *buffer, size_t len)
1644 data_len = hai->hai_len - sizeof(*hai);
1645 for (i = 0; (i < data_len) && (len > 2); i++) {
1646 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1656 /* Copytool action list */
1657 #define HAL_VERSION 1
1658 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1659 struct hsm_action_list {
1661 __u32 hal_count; /* number of hai's to follow */
1662 __u64 hal_compound_id; /* returned by coordinator */
1664 __u32 hal_archive_id; /* which archive backend */
1666 char hal_fsname[0]; /* null-terminated */
1667 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1668 boundaries. See hai_zero */
1669 } __attribute__((packed));
1671 #ifndef HAVE_CFS_SIZE_ROUND
1672 static inline int cfs_size_round (int val)
1674 return (val + 7) & (~0x7);
1676 #define HAVE_CFS_SIZE_ROUND
1679 /* Return pointer to first hai in action list */
1680 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1682 return (struct hsm_action_item *)(hal->hal_fsname +
1683 cfs_size_round(strlen(hal-> \
1687 /* Return pointer to next hai */
1688 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1690 return (struct hsm_action_item *)((char *)hai +
1691 cfs_size_round(hai->hai_len));
1694 /* Return size of an hsm_action_list */
1695 static inline size_t hal_size(struct hsm_action_list *hal)
1699 struct hsm_action_item *hai;
1701 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1702 hai = hai_first(hal);
1703 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1704 sz += cfs_size_round(hai->hai_len);
1710 * describe the attributes to be set on imported file
1712 struct hsm_user_import {
1721 __u32 hui_archive_id;
1724 /* Copytool progress reporting */
1725 #define HP_FLAG_COMPLETED 0x01
1726 #define HP_FLAG_RETRY 0x02
1728 struct hsm_progress {
1729 struct lu_fid hp_fid;
1731 struct hsm_extent hp_extent;
1733 __u16 hp_errval; /* positive val */
1738 __u64 hc_data_version;
1740 __u16 hc_errval; /* positive val */
1742 struct hsm_action_item hc_hai;
1746 enum llapi_json_types {
1747 LLAPI_JSON_INTEGER = 1,
1753 struct llapi_json_item {
1762 struct llapi_json_item *lji_next;
1765 struct llapi_json_item_list {
1766 int ljil_item_count;
1767 struct llapi_json_item *ljil_items;
1770 enum lu_ladvise_type {
1771 LU_LADVISE_INVALID = 0,
1772 LU_LADVISE_WILLREAD = 1,
1773 LU_LADVISE_DONTNEED = 2,
1774 LU_LADVISE_LOCKNOEXPAND = 3,
1775 LU_LADVISE_LOCKAHEAD = 4,
1779 #define LU_LADVISE_NAMES { \
1780 [LU_LADVISE_WILLREAD] = "willread", \
1781 [LU_LADVISE_DONTNEED] = "dontneed", \
1782 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
1783 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
1786 /* This is the userspace argument for ladvise. It is currently the same as
1787 * what goes on the wire (struct lu_ladvise), but is defined separately as we
1788 * may need info which is only used locally. */
1789 struct llapi_lu_ladvise {
1790 __u16 lla_advice; /* advice type */
1791 __u16 lla_value1; /* values for different advice types */
1793 __u64 lla_start; /* first byte of extent for advice */
1794 __u64 lla_end; /* last byte of extent for advice */
1800 LF_ASYNC = 0x00000001,
1801 LF_UNSET = 0x00000002,
1804 #define LADVISE_MAGIC 0x1ADF1CE0
1805 /* Masks of valid flags for each advice */
1806 #define LF_LOCKNOEXPAND_MASK LF_UNSET
1807 /* Flags valid for all advices not explicitly specified */
1808 #define LF_DEFAULT_MASK LF_ASYNC
1810 #define LF_MASK (LF_ASYNC | LF_UNSET)
1812 #define lla_lockahead_mode lla_value1
1813 #define lla_peradvice_flags lla_value2
1814 #define lla_lockahead_result lla_value3
1816 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
1817 * is used on the wire. It is defined separately as we may need info which is
1818 * only used locally. */
1819 struct llapi_ladvise_hdr {
1820 __u32 lah_magic; /* LADVISE_MAGIC */
1821 __u32 lah_count; /* number of advices */
1822 __u64 lah_flags; /* from enum ladvise_flag */
1823 __u32 lah_value1; /* unused */
1824 __u32 lah_value2; /* unused */
1825 __u64 lah_value3; /* unused */
1826 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
1829 #define LAH_COUNT_MAX (1024)
1833 SK_CRYPT_INVALID = -1,
1835 SK_CRYPT_AES256_CTR = 1,
1840 SK_HMAC_INVALID = -1,
1847 struct sk_crypt_type {
1852 struct sk_hmac_type {
1857 enum lock_mode_user {
1863 #define LOCK_MODE_NAMES { \
1864 [MODE_READ_USER] = "READ",\
1865 [MODE_WRITE_USER] = "WRITE"\
1868 enum lockahead_results {
1869 LLA_RESULT_SENT = 0,
1870 LLA_RESULT_DIFFERENT,
1874 #if defined(__cplusplus)
1878 /** @} lustreuser */
1880 #endif /* _LUSTRE_USER_H */