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/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,
337 LL_LEASE_LAYOUT_MERGE = 0x4,
340 #define IOC_IDS_MAX 4096
341 struct ll_ioc_lease {
349 * The ioctl naming rules:
350 * LL_* - works on the currently opened filehandle instead of parent dir
351 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
352 * *_MDC_* - gets/sets data related to MDC
353 * *_LOV_* - gets/sets data related to OSC/LOV
354 * *FILE* - called on parent dir and passes in a filename
355 * *STRIPE* - set/get lov_user_md
356 * *INFO - set/get lov_user_mds_data
358 /* lustre_ioctl.h 101-150 */
359 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
360 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
361 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
362 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
363 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
364 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
365 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
366 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
367 /* LL_IOC_RECREATE_OBJ 157 obsolete */
368 /* LL_IOC_RECREATE_FID 157 obsolete */
369 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
370 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
371 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
372 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
373 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
374 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
375 /* IOC_LOV_GETINFO 165 obsolete */
376 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
377 /* LL_IOC_RMTACL 167 obsolete */
378 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
379 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
380 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
381 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
382 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
383 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
384 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
385 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
386 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
387 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
388 /* lustre_ioctl.h 177-210 */
389 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
390 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
391 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
392 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
393 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
394 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
395 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
396 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
397 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
398 struct lustre_swap_layouts)
399 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
400 struct hsm_current_action)
401 /* lustre_ioctl.h 221-232 */
402 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
403 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
404 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
405 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
406 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
407 #define LL_IOC_GET_LEASE _IO('f', 244)
408 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
409 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
410 #define LL_IOC_MIGRATE _IOR('f', 247, int)
411 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
412 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
413 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
415 #ifndef FS_IOC_FSGETXATTR
417 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
420 __u32 fsx_xflags; /* xflags field value (get/set) */
421 __u32 fsx_extsize; /* extsize field value (get/set)*/
422 __u32 fsx_nextents; /* nextents field value (get) */
423 __u32 fsx_projid; /* project identifier (get/set) */
424 unsigned char fsx_pad[12];
426 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
427 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
429 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
430 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
431 #define LL_PROJINHERIT_FL 0x20000000
434 #define LL_STATFS_LMV 1
435 #define LL_STATFS_LOV 2
436 #define LL_STATFS_NODELAY 4
438 #define IOC_MDC_TYPE 'i'
439 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
440 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
441 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
442 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
444 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
446 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
447 * files, but are unlikely to be used in practice and are not harmful if
448 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
449 * devices and are safe for use on new files. See LU-4209. */
450 /* To be compatible with old statically linked binary we keep the check for
451 * the older 0100000000 flag. This is already removed upstream. LU-812. */
452 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
453 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
454 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
455 O_LOV_DELAY_CREATE_MASK)
457 #define LL_FILE_IGNORE_LOCK 0x00000001
458 #define LL_FILE_GROUP_LOCKED 0x00000002
459 #define LL_FILE_READAHEA 0x00000004
460 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
461 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
463 #define LOV_USER_MAGIC_V1 0x0BD10BD0
464 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
465 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
466 #define LOV_USER_MAGIC_V3 0x0BD30BD0
467 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
468 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
469 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
471 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
472 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
473 #define LMV_USER_MAGIC_SPECIFIC 0x0CD40CD0
475 #define LOV_PATTERN_NONE 0x000
476 #define LOV_PATTERN_RAID0 0x001
477 #define LOV_PATTERN_RAID1 0x002
478 #define LOV_PATTERN_MDT 0x100
479 #define LOV_PATTERN_CMOBD 0x200
481 #define LOV_PATTERN_F_MASK 0xffff0000
482 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
483 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
484 #define LOV_PATTERN_DEFAULT 0xffffffff
486 static inline bool lov_pattern_supported(__u32 pattern)
488 return pattern == LOV_PATTERN_RAID0 ||
489 pattern == LOV_PATTERN_MDT ||
490 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_F_RELEASED);
493 #define LOV_MAXPOOLNAME 15
494 #define LOV_POOLNAMEF "%.15s"
496 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
497 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
498 #define LOV_MAX_STRIPE_COUNT_OLD 160
499 /* This calculation is crafted so that input of 4096 will result in 160
500 * which in turn is equal to old maximal stripe count.
501 * XXX: In fact this is too simpified for now, what it also need is to get
502 * ea_type argument to clearly know how much space each stripe consumes.
504 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
505 * allocation that is sufficient for the current generation of systems.
507 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
508 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
509 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
510 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
512 #define XATTR_LUSTRE_PREFIX "lustre."
513 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
515 #define lov_user_ost_data lov_user_ost_data_v1
516 struct lov_user_ost_data_v1 { /* per-stripe data structure */
517 struct ost_id l_ost_oi; /* OST object ID */
518 __u32 l_ost_gen; /* generation of this OST index */
519 __u32 l_ost_idx; /* OST index in LOV */
520 } __attribute__((packed));
522 #define lov_user_md lov_user_md_v1
523 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
524 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
525 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
526 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
527 __u32 lmm_stripe_size; /* size of stripe in bytes */
528 __u16 lmm_stripe_count; /* num stripes in use for this object */
530 __u16 lmm_stripe_offset; /* starting stripe offset in
531 * lmm_objects, use when writing */
532 __u16 lmm_layout_gen; /* layout generation number
533 * used when reading */
535 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
536 } __attribute__((packed, __may_alias__));
538 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
539 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
540 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
541 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
542 __u32 lmm_stripe_size; /* size of stripe in bytes */
543 __u16 lmm_stripe_count; /* num stripes in use for this object */
545 __u16 lmm_stripe_offset; /* starting stripe offset in
546 * lmm_objects, use when writing */
547 __u16 lmm_layout_gen; /* layout generation number
548 * used when reading */
550 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
551 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
552 } __attribute__((packed));
559 #define DEXT "[%#llx, %#llx)"
560 #define PEXT(ext) (ext)->e_start, (ext)->e_end
562 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
563 struct lu_extent *e2)
565 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
568 static inline bool lu_extent_is_whole(struct lu_extent *e)
570 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
573 enum lov_comp_md_entry_flags {
574 LCME_FL_STALE = 0x00000001, /* FLR: stale data */
575 LCME_FL_PREF_RD = 0x00000002, /* FLR: preferred for reading */
576 LCME_FL_PREF_WR = 0x00000004, /* FLR: preferred for writing */
577 LCME_FL_PREF_RW = LCME_FL_PREF_RD | LCME_FL_PREF_WR,
578 LCME_FL_OFFLINE = 0x00000008, /* Not used */
579 LCME_FL_INIT = 0x00000010, /* instantiated */
580 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
581 won't be stored on disk */
584 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT | LCME_FL_STALE | \
586 /* The flags can be set by users at mirror creation time. */
587 #define LCME_USER_FLAGS (LCME_FL_PREF_RW)
589 /* the highest bit in obdo::o_layout_version is used to mark if the file is
591 #define LU_LAYOUT_RESYNC LCME_FL_NEG
593 /* lcme_id can be specified as certain flags, and the the first
594 * bit of lcme_id is used to indicate that the ID is representing
595 * certain LCME_FL_* but not a real ID. Which implies we can have
596 * at most 31 flags (see LCME_FL_XXX). */
599 LCME_ID_MAX = 0x7FFFFFFF,
600 LCME_ID_ALL = 0xFFFFFFFF,
601 LCME_ID_NOT_ID = LCME_FL_NEG
604 #define LCME_ID_MASK LCME_ID_MAX
606 struct lov_comp_md_entry_v1 {
607 __u32 lcme_id; /* unique id of component */
608 __u32 lcme_flags; /* LCME_FL_XXX */
609 struct lu_extent lcme_extent; /* file extent for component */
610 __u32 lcme_offset; /* offset of component blob,
611 start from lov_comp_md_v1 */
612 __u32 lcme_size; /* size of component blob */
613 __u64 lcme_padding[2];
614 } __attribute__((packed));
616 #define SEQ_ID_MAX 0x0000FFFF
617 #define SEQ_ID_MASK SEQ_ID_MAX
618 /* bit 30:16 of lcme_id is used to store mirror id */
619 #define MIRROR_ID_MASK 0x7FFF0000
620 #define MIRROR_ID_SHIFT 16
622 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
624 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
627 static inline __u16 mirror_id_of(__u32 id)
629 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
633 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
635 enum lov_comp_md_flags {
636 /* the least 2 bits are used by FLR to record file state */
639 LCM_FL_WRITE_PENDING = 2,
640 LCM_FL_SYNC_PENDING = 3,
641 LCM_FL_FLR_MASK = 0x3,
644 struct lov_comp_md_v1 {
645 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
646 __u32 lcm_size; /* overall size including this struct */
647 __u32 lcm_layout_gen;
649 __u16 lcm_entry_count;
650 /* lcm_mirror_count stores the number of actual mirrors minus 1,
651 * so that non-flr files will have value 0 meaning 1 mirror. */
652 __u16 lcm_mirror_count;
653 __u16 lcm_padding1[3];
655 struct lov_comp_md_entry_v1 lcm_entries[0];
656 } __attribute__((packed));
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)
946 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
947 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
948 struct lustre_swap_layouts {
956 /** Bit-mask of valid attributes */
957 /* The LA_* flags are written to disk as part of the ChangeLog records
958 * so they are part of the on-disk and network protocol, and cannot be changed.
959 * Only the first 12 bits are currently saved.
974 LA_BLKSIZE = 1 << 12,
975 LA_KILL_SUID = 1 << 13,
976 LA_KILL_SGID = 1 << 14,
978 LA_LAYOUT_VERSION = 1 << 16,
980 * Attributes must be transmitted to OST objects
982 LA_REMOTE_ATTR_SET = (LA_UID | LA_GID | LA_PROJID | LA_LAYOUT_VERSION)
985 /********* Changelogs **********/
986 /** Changelog record types */
987 enum changelog_rec_type {
989 CL_CREATE = 1, /* namespace */
990 CL_MKDIR = 2, /* namespace */
991 CL_HARDLINK = 3, /* namespace */
992 CL_SOFTLINK = 4, /* namespace */
993 CL_MKNOD = 5, /* namespace */
994 CL_UNLINK = 6, /* namespace */
995 CL_RMDIR = 7, /* namespace */
996 CL_RENAME = 8, /* namespace */
997 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
998 CL_OPEN = 10, /* not currently used */
999 CL_CLOSE = 11, /* may be written to log only with mtime change */
1000 CL_LAYOUT = 12, /* file layout/striping modified */
1004 CL_HSM = 16, /* HSM specific events, see flags */
1005 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
1009 CL_FLRW = 21, /* FLR: file was firstly written */
1010 CL_RESYNC = 22, /* FLR: file was resync-ed */
1014 static inline const char *changelog_type2str(int type) {
1015 static const char *changelog_str[] = {
1016 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
1017 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
1018 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
1022 if (type >= 0 && type < CL_LAST)
1023 return changelog_str[type];
1027 /* per-record flags */
1028 #define CLF_FLAGSHIFT 12
1029 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
1030 #define CLF_VERMASK (~CLF_FLAGMASK)
1031 enum changelog_rec_flags {
1032 CLF_VERSION = 0x1000,
1033 CLF_RENAME = 0x2000,
1035 CLF_EXTRA_FLAGS = 0x8000,
1036 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS
1040 /* Anything under the flagmask may be per-type (if desired) */
1041 /* Flags for unlink */
1042 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1043 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1044 /* HSM cleaning needed */
1045 /* Flags for rename */
1046 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1048 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1049 * has an archive in backend */
1052 /* 12b used (from high weight to low weight):
1057 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1058 #define CLF_HSM_ERR_H 6
1059 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1060 #define CLF_HSM_EVENT_H 9
1061 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1062 #define CLF_HSM_FLAG_H 11
1063 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1064 #define CLF_HSM_SPARE_H 15
1065 #define CLF_HSM_LAST 15
1067 /* Remove bits higher than _h, then extract the value
1068 * between _h and _l by shifting lower weigth to bit 0. */
1069 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1070 >> (CLF_HSM_LAST - _h + _l))
1072 #define CLF_HSM_SUCCESS 0x00
1073 #define CLF_HSM_MAXERROR 0x7E
1074 #define CLF_HSM_ERROVERFLOW 0x7F
1076 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1078 /* 3 bits field => 8 values allowed */
1090 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1092 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1096 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
1098 *flags |= (he << CLF_HSM_EVENT_L);
1101 static inline __u16 hsm_get_cl_flags(int flags)
1103 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1106 static inline void hsm_set_cl_flags(int *flags, int bits)
1108 *flags |= (bits << CLF_HSM_FLAG_L);
1111 static inline int hsm_get_cl_error(int flags)
1113 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1116 static inline void hsm_set_cl_error(int *flags, int error)
1118 *flags |= (error << CLF_HSM_ERR_L);
1121 enum changelog_rec_extra_flags {
1123 CLFE_UIDGID = 0x0001,
1125 CLFE_SUPPORTED = CLFE_UIDGID | CLFE_NID
1128 enum changelog_send_flag {
1129 /* Not yet implemented */
1130 CHANGELOG_FLAG_FOLLOW = 0x01,
1131 /* Blocking IO makes sense in case of slow user parsing of the records,
1132 * but it also prevents us from cleaning up if the records are not
1134 CHANGELOG_FLAG_BLOCK = 0x02,
1135 /* Pack jobid into the changelog records if available. */
1136 CHANGELOG_FLAG_JOBID = 0x04,
1137 /* Pack additional flag bits into the changelog record */
1138 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1141 enum changelog_send_extra_flag {
1142 /* Pack uid/gid into the changelog record */
1143 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1144 /* Pack nid into the changelog record */
1145 CHANGELOG_EXTRA_FLAG_NID = 0x02,
1148 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
1149 changelog_rec_offset(CLF_SUPPORTED, \
1152 /* 31 usable bytes string + null terminator. */
1153 #define LUSTRE_JOBID_SIZE 32
1155 /* This is the minimal changelog record. It can contain extensions
1156 * such as rename fields or process jobid. Its exact content is described
1157 * by the cr_flags and cr_extra_flags.
1159 * Extensions are packed in the same order as their corresponding flags,
1160 * then in the same order as their corresponding extra flags.
1162 struct changelog_rec {
1164 __u16 cr_flags; /**< \a changelog_rec_flags */
1165 __u32 cr_type; /**< \a changelog_rec_type */
1166 __u64 cr_index; /**< changelog record number */
1167 __u64 cr_prev; /**< last index for this target fid */
1170 struct lu_fid cr_tfid; /**< target fid */
1171 __u32 cr_markerflags; /**< CL_MARK flags */
1173 struct lu_fid cr_pfid; /**< parent fid */
1176 /* Changelog extension for RENAME. */
1177 struct changelog_ext_rename {
1178 struct lu_fid cr_sfid; /**< source fid, or zero */
1179 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1182 /* Changelog extension to include JOBID. */
1183 struct changelog_ext_jobid {
1184 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1187 /* Changelog extension to include additional flags. */
1188 struct changelog_ext_extra_flags {
1189 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1192 /* Changelog extra extension to include UID/GID. */
1193 struct changelog_ext_uidgid {
1198 /* Changelog extra extension to include NID. */
1199 struct changelog_ext_nid {
1200 /* have __u64 instead of lnet_nid_t type for use by client api */
1202 /* for use when IPv6 support is added */
1207 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1208 const struct changelog_rec *rec);
1210 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1211 enum changelog_rec_extra_flags cref)
1213 size_t size = sizeof(struct changelog_rec);
1215 if (crf & CLF_RENAME)
1216 size += sizeof(struct changelog_ext_rename);
1218 if (crf & CLF_JOBID)
1219 size += sizeof(struct changelog_ext_jobid);
1221 if (crf & CLF_EXTRA_FLAGS) {
1222 size += sizeof(struct changelog_ext_extra_flags);
1223 if (cref & CLFE_UIDGID)
1224 size += sizeof(struct changelog_ext_uidgid);
1225 if (cref & CLFE_NID)
1226 size += sizeof(struct changelog_ext_nid);
1232 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1234 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1236 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1237 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1239 return changelog_rec_offset(rec->cr_flags, cref);
1242 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1244 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1248 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1250 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1252 return (struct changelog_ext_rename *)((char *)rec +
1253 changelog_rec_offset(crf,
1257 /* The jobid follows the rename extension, if present */
1259 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1261 enum changelog_rec_flags crf = rec->cr_flags &
1262 (CLF_VERSION | CLF_RENAME);
1264 return (struct changelog_ext_jobid *)((char *)rec +
1265 changelog_rec_offset(crf,
1269 /* The additional flags follow the rename and jobid extensions, if present */
1271 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1272 const struct changelog_rec *rec)
1274 enum changelog_rec_flags crf = rec->cr_flags &
1275 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1277 return (struct changelog_ext_extra_flags *)((char *)rec +
1278 changelog_rec_offset(crf,
1282 /* The uid/gid is the first extra extension */
1284 struct changelog_ext_uidgid *changelog_rec_uidgid(
1285 const struct changelog_rec *rec)
1287 enum changelog_rec_flags crf = rec->cr_flags &
1288 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1290 return (struct changelog_ext_uidgid *)((char *)rec +
1291 changelog_rec_offset(crf,
1295 /* The nid is the second extra extension */
1297 struct changelog_ext_nid *changelog_rec_nid(const struct changelog_rec *rec)
1299 enum changelog_rec_flags crf = rec->cr_flags &
1300 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1301 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1303 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1304 cref = changelog_rec_extra_flags(rec)->cr_extra_flags &
1307 return (struct changelog_ext_nid *)((char *)rec +
1308 changelog_rec_offset(crf, cref));
1311 /* The name follows the rename, jobid and extra flags extns, if present */
1312 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1314 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1316 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1317 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1319 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1320 cref & CLFE_SUPPORTED);
1323 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1325 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1328 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1330 char *cr_name = changelog_rec_name(rec);
1332 return cr_name + strlen(cr_name) + 1;
1336 * Remap a record to the desired format as specified by the crf flags.
1337 * The record must be big enough to contain the final remapped version.
1338 * Superfluous extension fields are removed and missing ones are added
1339 * and zeroed. The flags of the record are updated accordingly.
1341 * The jobid and rename extensions can be added to a record, to match the
1342 * format an application expects, typically. In this case, the newly added
1343 * fields will be zeroed.
1344 * The Jobid field can be removed, to guarantee compatibility with older
1345 * clients that don't expect this field in the records they process.
1347 * The following assumptions are being made:
1348 * - CLF_RENAME will not be removed
1349 * - CLF_JOBID will not be added without CLF_RENAME being added too
1350 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1352 * @param[in,out] rec The record to remap.
1353 * @param[in] crf_wanted Flags describing the desired extensions.
1354 * @param[in] cref_want Flags describing the desired extra extensions.
1356 static inline void changelog_remap_rec(struct changelog_rec *rec,
1357 enum changelog_rec_flags crf_wanted,
1358 enum changelog_rec_extra_flags cref_want)
1360 char *nid_mov = NULL;
1361 char *uidgid_mov = NULL;
1365 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1367 crf_wanted &= CLF_SUPPORTED;
1368 cref_want &= CLFE_SUPPORTED;
1370 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1371 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1372 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1373 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1379 /* First move the variable-length name field */
1380 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1381 changelog_rec_name(rec), rec->cr_namelen);
1383 /* Locations of extensions in the remapped record */
1384 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1385 nid_mov = (char *)rec +
1386 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1387 cref_want & ~CLFE_NID);
1388 uidgid_mov = (char *)rec +
1389 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1390 cref_want & ~(CLFE_UIDGID |
1392 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1395 ef_mov = (char *)rec +
1396 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1398 jid_mov = (char *)rec +
1399 changelog_rec_offset(crf_wanted &
1400 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1402 rnm_mov = (char *)rec +
1403 changelog_rec_offset(crf_wanted &
1409 /* Move the extension fields to the desired positions */
1410 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1411 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1412 if ((cref_want & CLFE_NID) && (cref & CLFE_NID))
1413 memmove(nid_mov, changelog_rec_nid(rec),
1414 sizeof(struct changelog_ext_nid));
1416 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1417 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1418 sizeof(struct changelog_ext_uidgid));
1420 memmove(ef_mov, changelog_rec_extra_flags(rec),
1421 sizeof(struct changelog_ext_extra_flags));
1424 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1425 memmove(jid_mov, changelog_rec_jobid(rec),
1426 sizeof(struct changelog_ext_jobid));
1428 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1429 memmove(rnm_mov, changelog_rec_rename(rec),
1430 sizeof(struct changelog_ext_rename));
1432 /* Clear newly added fields */
1433 if (nid_mov && (cref_want & CLFE_NID) &&
1435 memset(nid_mov, 0, sizeof(struct changelog_ext_nid));
1437 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1438 !(cref & CLFE_UIDGID))
1439 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1441 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1442 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1443 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1445 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1446 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1448 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1449 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1451 /* Update the record's flags accordingly */
1452 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1453 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1454 changelog_rec_extra_flags(rec)->cr_extra_flags =
1455 changelog_rec_extra_flags(rec)->cr_extra_flags |
1459 enum changelog_message_type {
1460 CL_RECORD = 10, /* message is a changelog_rec */
1461 CL_EOF = 11, /* at end of current changelog */
1464 /********* Misc **********/
1466 struct ioc_data_version {
1468 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1469 __u32 idv_flags; /* enum ioc_data_version_flags */
1472 enum ioc_data_version_flags {
1473 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1474 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1478 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1481 #define dot_lustre_name ".lustre"
1484 /********* HSM **********/
1486 /** HSM per-file state
1487 * See HSM_FLAGS below.
1490 HS_NONE = 0x00000000,
1491 HS_EXISTS = 0x00000001,
1492 HS_DIRTY = 0x00000002,
1493 HS_RELEASED = 0x00000004,
1494 HS_ARCHIVED = 0x00000008,
1495 HS_NORELEASE = 0x00000010,
1496 HS_NOARCHIVE = 0x00000020,
1497 HS_LOST = 0x00000040,
1500 /* HSM user-setable flags. */
1501 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1503 /* Other HSM flags. */
1504 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1507 * All HSM-related possible flags that could be applied to a file.
1508 * This should be kept in sync with hsm_states.
1510 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1513 * HSMÂ request progress state
1515 enum hsm_progress_states {
1522 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1525 case HPS_WAITING: return "waiting";
1526 case HPS_RUNNING: return "running";
1527 case HPS_DONE: return "done";
1528 default: return "unknown";
1535 } __attribute__((packed));
1538 * Current HSM states of a Lustre file.
1540 * This structure purpose is to be sent to user-space mainly. It describes the
1541 * current HSM flags and in-progress action.
1543 struct hsm_user_state {
1544 /** Current HSM states, from enum hsm_states. */
1546 __u32 hus_archive_id;
1547 /** The current undergoing action, if there is one */
1548 __u32 hus_in_progress_state;
1549 __u32 hus_in_progress_action;
1550 struct hsm_extent hus_in_progress_location;
1551 char hus_extended_info[];
1554 struct hsm_state_set_ioc {
1555 struct lu_fid hssi_fid;
1557 __u64 hssi_clearmask;
1561 * This structure describes the current in-progress action for a file.
1562 * it is retuned to user space and send over the wire
1564 struct hsm_current_action {
1565 /** The current undergoing action, if there is one */
1566 /* state is one of hsm_progress_states */
1568 /* action is one of hsm_user_action */
1570 struct hsm_extent hca_location;
1573 /***** HSM user requests ******/
1574 /* User-generated (lfs/ioctl) request types */
1575 enum hsm_user_action {
1576 HUA_NONE = 1, /* no action (noop) */
1577 HUA_ARCHIVE = 10, /* copy to hsm */
1578 HUA_RESTORE = 11, /* prestage */
1579 HUA_RELEASE = 12, /* drop ost objects */
1580 HUA_REMOVE = 13, /* remove from archive */
1581 HUA_CANCEL = 14 /* cancel a request */
1584 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1587 case HUA_NONE: return "NOOP";
1588 case HUA_ARCHIVE: return "ARCHIVE";
1589 case HUA_RESTORE: return "RESTORE";
1590 case HUA_RELEASE: return "RELEASE";
1591 case HUA_REMOVE: return "REMOVE";
1592 case HUA_CANCEL: return "CANCEL";
1593 default: return "UNKNOWN";
1598 * List of hr_flags (bit field)
1600 #define HSM_FORCE_ACTION 0x0001
1601 /* used by CT, cannot be set by user */
1602 #define HSM_GHOST_COPY 0x0002
1605 * Contains all the fixed part of struct hsm_user_request.
1608 struct hsm_request {
1609 __u32 hr_action; /* enum hsm_user_action */
1610 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1611 __u64 hr_flags; /* request flags */
1612 __u32 hr_itemcount; /* item count in hur_user_item vector */
1616 struct hsm_user_item {
1617 struct lu_fid hui_fid;
1618 struct hsm_extent hui_extent;
1619 } __attribute__((packed));
1621 struct hsm_user_request {
1622 struct hsm_request hur_request;
1623 struct hsm_user_item hur_user_item[0];
1624 /* extra data blob at end of struct (after all
1625 * hur_user_items), only use helpers to access it
1627 } __attribute__((packed));
1629 /** Return pointer to data field in a hsm user request */
1630 static inline void *hur_data(struct hsm_user_request *hur)
1632 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1636 * Compute the current length of the provided hsm_user_request. This returns -1
1637 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1639 * return -1 on bounds check error.
1641 static inline ssize_t hur_len(struct hsm_user_request *hur)
1645 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1646 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1647 (__u64)hur->hur_request.hr_itemcount *
1648 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1650 if (size != (ssize_t)size)
1656 /****** HSM RPCs to copytool *****/
1657 /* Message types the copytool may receive */
1658 enum hsm_message_type {
1659 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1662 /* Actions the copytool may be instructed to take for a given action_item */
1663 enum hsm_copytool_action {
1664 HSMA_NONE = 10, /* no action */
1665 HSMA_ARCHIVE = 20, /* arbitrary offset */
1671 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1674 case HSMA_NONE: return "NOOP";
1675 case HSMA_ARCHIVE: return "ARCHIVE";
1676 case HSMA_RESTORE: return "RESTORE";
1677 case HSMA_REMOVE: return "REMOVE";
1678 case HSMA_CANCEL: return "CANCEL";
1679 default: return "UNKNOWN";
1683 /* Copytool item action description */
1684 struct hsm_action_item {
1685 __u32 hai_len; /* valid size of this struct */
1686 __u32 hai_action; /* hsm_copytool_action, but use known size */
1687 struct lu_fid hai_fid; /* Lustre FID to operate on */
1688 struct lu_fid hai_dfid; /* fid used for data access */
1689 struct hsm_extent hai_extent; /* byte range to operate on */
1690 __u64 hai_cookie; /* action cookie from coordinator */
1691 __u64 hai_gid; /* grouplock id */
1692 char hai_data[0]; /* variable length */
1693 } __attribute__((packed));
1696 * helper function which print in hexa the first bytes of
1699 * \param hai [IN] record to print
1700 * \param buffer [IN,OUT] buffer to write the hex string to
1701 * \param len [IN] max buffer length
1705 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1706 char *buffer, size_t len)
1713 data_len = hai->hai_len - sizeof(*hai);
1714 for (i = 0; (i < data_len) && (len > 2); i++) {
1715 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1725 /* Copytool action list */
1726 #define HAL_VERSION 1
1727 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1728 struct hsm_action_list {
1730 __u32 hal_count; /* number of hai's to follow */
1731 __u64 hal_compound_id; /* returned by coordinator */
1733 __u32 hal_archive_id; /* which archive backend */
1735 char hal_fsname[0]; /* null-terminated */
1736 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1737 boundaries. See hai_zero */
1738 } __attribute__((packed));
1740 #ifndef HAVE_CFS_SIZE_ROUND
1741 static inline int cfs_size_round (int val)
1743 return (val + 7) & (~0x7);
1745 #define HAVE_CFS_SIZE_ROUND
1748 /* Return pointer to first hai in action list */
1749 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1751 return (struct hsm_action_item *)(hal->hal_fsname +
1752 cfs_size_round(strlen(hal-> \
1756 /* Return pointer to next hai */
1757 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1759 return (struct hsm_action_item *)((char *)hai +
1760 cfs_size_round(hai->hai_len));
1763 /* Return size of an hsm_action_list */
1764 static inline size_t hal_size(struct hsm_action_list *hal)
1768 struct hsm_action_item *hai;
1770 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1771 hai = hai_first(hal);
1772 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1773 sz += cfs_size_round(hai->hai_len);
1779 * describe the attributes to be set on imported file
1781 struct hsm_user_import {
1790 __u32 hui_archive_id;
1793 /* Copytool progress reporting */
1794 #define HP_FLAG_COMPLETED 0x01
1795 #define HP_FLAG_RETRY 0x02
1797 struct hsm_progress {
1798 struct lu_fid hp_fid;
1800 struct hsm_extent hp_extent;
1802 __u16 hp_errval; /* positive val */
1807 __u64 hc_data_version;
1809 __u16 hc_errval; /* positive val */
1811 struct hsm_action_item hc_hai;
1815 enum llapi_json_types {
1816 LLAPI_JSON_INTEGER = 1,
1822 struct llapi_json_item {
1831 struct llapi_json_item *lji_next;
1834 struct llapi_json_item_list {
1835 int ljil_item_count;
1836 struct llapi_json_item *ljil_items;
1839 enum lu_ladvise_type {
1840 LU_LADVISE_INVALID = 0,
1841 LU_LADVISE_WILLREAD = 1,
1842 LU_LADVISE_DONTNEED = 2,
1843 LU_LADVISE_LOCKNOEXPAND = 3,
1844 LU_LADVISE_LOCKAHEAD = 4,
1848 #define LU_LADVISE_NAMES { \
1849 [LU_LADVISE_WILLREAD] = "willread", \
1850 [LU_LADVISE_DONTNEED] = "dontneed", \
1851 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
1852 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
1855 /* This is the userspace argument for ladvise. It is currently the same as
1856 * what goes on the wire (struct lu_ladvise), but is defined separately as we
1857 * may need info which is only used locally. */
1858 struct llapi_lu_ladvise {
1859 __u16 lla_advice; /* advice type */
1860 __u16 lla_value1; /* values for different advice types */
1862 __u64 lla_start; /* first byte of extent for advice */
1863 __u64 lla_end; /* last byte of extent for advice */
1869 LF_ASYNC = 0x00000001,
1870 LF_UNSET = 0x00000002,
1873 #define LADVISE_MAGIC 0x1ADF1CE0
1874 /* Masks of valid flags for each advice */
1875 #define LF_LOCKNOEXPAND_MASK LF_UNSET
1876 /* Flags valid for all advices not explicitly specified */
1877 #define LF_DEFAULT_MASK LF_ASYNC
1879 #define LF_MASK (LF_ASYNC | LF_UNSET)
1881 #define lla_lockahead_mode lla_value1
1882 #define lla_peradvice_flags lla_value2
1883 #define lla_lockahead_result lla_value3
1885 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
1886 * is used on the wire. It is defined separately as we may need info which is
1887 * only used locally. */
1888 struct llapi_ladvise_hdr {
1889 __u32 lah_magic; /* LADVISE_MAGIC */
1890 __u32 lah_count; /* number of advices */
1891 __u64 lah_flags; /* from enum ladvise_flag */
1892 __u32 lah_value1; /* unused */
1893 __u32 lah_value2; /* unused */
1894 __u64 lah_value3; /* unused */
1895 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
1898 #define LAH_COUNT_MAX (1024)
1902 SK_CRYPT_INVALID = -1,
1904 SK_CRYPT_AES256_CTR = 1,
1909 SK_HMAC_INVALID = -1,
1916 struct sk_crypt_type {
1921 struct sk_hmac_type {
1926 enum lock_mode_user {
1932 #define LOCK_MODE_NAMES { \
1933 [MODE_READ_USER] = "READ",\
1934 [MODE_WRITE_USER] = "WRITE"\
1937 enum lockahead_results {
1938 LLA_RESULT_SENT = 0,
1939 LLA_RESULT_DIFFERENT,
1943 #if defined(__cplusplus)
1947 /** @} lustreuser */
1949 #endif /* _LUSTRE_USER_H */