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 */
134 __u32 os_state; /**< obd_statfs_state OS_STATE_* flag */
135 __u32 os_fprecreated; /* objs available now to the caller */
136 /* used in QoS code to find preferred
151 * FID is a cluster-wide unique identifier of a file or an object (stripe).
152 * FIDs are never reused.
156 * FID sequence. Sequence is a unit of migration: all files (objects)
157 * with FIDs from a given sequence are stored on the same server.
158 * Lustre should support 2^64 objects, so even if each sequence
159 * has only a single object we can still enumerate 2^64 objects.
162 /* FID number within sequence. */
165 * FID version, used to distinguish different versions (in the sense
166 * of snapshots, etc.) of the same file system object. Not currently
172 static inline bool fid_is_zero(const struct lu_fid *fid)
174 return fid->f_seq == 0 && fid->f_oid == 0;
177 /* Currently, the filter_fid::ff_parent::f_ver is not the real parent
178 * MDT-object's FID::f_ver, instead it is the OST-object index in its
179 * parent MDT-object's layout EA. */
180 #define f_stripe_idx f_ver
183 __u32 ol_stripe_size;
184 __u32 ol_stripe_count;
188 } __attribute__((packed));
190 /* keep this one for compatibility */
191 struct filter_fid_old {
192 struct lu_fid ff_parent;
198 struct lu_fid ff_parent;
199 struct ost_layout ff_layout;
200 __u32 ff_layout_version;
201 __u32 ff_range; /* range of layout version that
202 * write are allowed */
203 } __attribute__((packed));
205 /* Userspace should treat lu_fid as opaque, and only use the following methods
206 * to print or parse them. Other functions (e.g. compare, swab) could be moved
207 * here from lustre_idl.h if needed. */
211 LMAC_HSM = 0x00000001,
212 /* LMAC_SOM = 0x00000002, obsolete since 2.8.0 */
213 LMAC_NOT_IN_OI = 0x00000004, /* the object does NOT need OI mapping */
214 LMAC_FID_ON_OST = 0x00000008, /* For OST-object, its OI mapping is
215 * under /O/<seq>/d<x>. */
216 LMAC_STRIPE_INFO = 0x00000010, /* stripe info in the LMA EA. */
217 LMAC_COMP_INFO = 0x00000020, /* Component info in the LMA EA. */
221 * Masks for all features that should be supported by a Lustre version to
222 * access a specific file.
223 * This information is stored in lustre_mdt_attrs::lma_incompat.
226 LMAI_RELEASED = 0x00000001, /* file is released */
227 LMAI_AGENT = 0x00000002, /* agent inode */
228 LMAI_REMOTE_PARENT = 0x00000004, /* the parent of the object
229 is on the remote MDT */
230 LMAI_STRIPED = 0x00000008, /* striped directory inode */
231 LMAI_ORPHAN = 0x00000010, /* inode is orphan */
232 LMA_INCOMPAT_SUPP = (LMAI_AGENT | LMAI_REMOTE_PARENT | \
233 LMAI_STRIPED | LMAI_ORPHAN)
238 * Following struct for object attributes, that will be kept inode's EA.
239 * Introduced in 2.0 release (please see b15993, for details)
240 * Added to all objects since Lustre 2.4 as contains self FID
242 struct lustre_mdt_attrs {
244 * Bitfield for supported data in this structure. From enum lma_compat.
245 * lma_self_fid and lma_flags are always available.
249 * Per-file incompat feature list. Lustre version should support all
250 * flags set in this field. The supported feature mask is available in
254 /** FID of this inode */
255 struct lu_fid lma_self_fid;
258 struct lustre_ost_attrs {
259 /* Use lustre_mdt_attrs directly for now, need a common header
260 * structure if want to change lustre_mdt_attrs in future. */
261 struct lustre_mdt_attrs loa_lma;
263 /* Below five elements are for OST-object's PFID EA, the
264 * lma_parent_fid::f_ver is composed of the stripe_count (high 16 bits)
265 * and the stripe_index (low 16 bits), the size should not exceed
266 * 5 * sizeof(__u64)) to be accessable by old Lustre. If the flag
267 * LMAC_STRIPE_INFO is set, then loa_parent_fid and loa_stripe_size
268 * are valid; if the flag LMAC_COMP_INFO is set, then the next three
269 * loa_comp_* elements are valid. */
270 struct lu_fid loa_parent_fid;
271 __u32 loa_stripe_size;
273 __u64 loa_comp_start;
278 * Prior to 2.4, the LMA structure also included SOM attributes which has since
279 * been moved to a dedicated xattr
280 * lma_flags was also removed because of lma_compat/incompat fields.
282 #define LMA_OLD_SIZE (sizeof(struct lustre_mdt_attrs) + 5 * sizeof(__u64))
285 LSOM_FL_VALID = 1 << 0,
288 struct lustre_som_attrs {
290 __u16 lsa_reserved[3];
296 * OST object IDentifier.
304 struct lu_fid oi_fid;
308 #define DOSTID "%#llx:%llu"
309 #define POSTID(oi) ((unsigned long long)ostid_seq(oi)), \
310 ((unsigned long long)ostid_id(oi))
312 struct ll_futimes_3 {
314 __u64 lfu_atime_nsec;
316 __u64 lfu_mtime_nsec;
318 __u64 lfu_ctime_nsec;
322 * Maximum number of mirrors currently implemented.
324 #define LUSTRE_MIRROR_COUNT_MAX 16
326 /* Lease types for use as arg and return of LL_IOC_{GET,SET}_LEASE ioctl. */
328 LL_LEASE_RDLCK = 0x01,
329 LL_LEASE_WRLCK = 0x02,
330 LL_LEASE_UNLCK = 0x04,
333 enum ll_lease_flags {
334 LL_LEASE_RESYNC = 0x1,
335 LL_LEASE_RESYNC_DONE = 0x2,
338 #define IOC_IDS_MAX 4096
339 struct ll_ioc_lease {
347 * The ioctl naming rules:
348 * LL_* - works on the currently opened filehandle instead of parent dir
349 * *_OBD_* - gets data for both OSC or MDC (LOV, LMV indirectly)
350 * *_MDC_* - gets/sets data related to MDC
351 * *_LOV_* - gets/sets data related to OSC/LOV
352 * *FILE* - called on parent dir and passes in a filename
353 * *STRIPE* - set/get lov_user_md
354 * *INFO - set/get lov_user_mds_data
356 /* lustre_ioctl.h 101-150 */
357 #define LL_IOC_GETFLAGS _IOR ('f', 151, long)
358 #define LL_IOC_SETFLAGS _IOW ('f', 152, long)
359 #define LL_IOC_CLRFLAGS _IOW ('f', 153, long)
360 #define LL_IOC_LOV_SETSTRIPE _IOW ('f', 154, long)
361 #define LL_IOC_LOV_SETSTRIPE_NEW _IOWR('f', 154, struct lov_user_md)
362 #define LL_IOC_LOV_GETSTRIPE _IOW ('f', 155, long)
363 #define LL_IOC_LOV_GETSTRIPE_NEW _IOR('f', 155, struct lov_user_md)
364 #define LL_IOC_LOV_SETEA _IOW ('f', 156, long)
365 /* LL_IOC_RECREATE_OBJ 157 obsolete */
366 /* LL_IOC_RECREATE_FID 157 obsolete */
367 #define LL_IOC_GROUP_LOCK _IOW ('f', 158, long)
368 #define LL_IOC_GROUP_UNLOCK _IOW ('f', 159, long)
369 /* LL_IOC_QUOTACHECK 160 OBD_IOC_QUOTACHECK */
370 /* LL_IOC_POLL_QUOTACHECK 161 OBD_IOC_POLL_QUOTACHECK */
371 /* LL_IOC_QUOTACTL 162 OBD_IOC_QUOTACTL */
372 #define IOC_OBD_STATFS _IOWR('f', 164, struct obd_statfs *)
373 /* IOC_LOV_GETINFO 165 obsolete */
374 #define LL_IOC_FLUSHCTX _IOW ('f', 166, long)
375 /* LL_IOC_RMTACL 167 obsolete */
376 #define LL_IOC_GETOBDCOUNT _IOR ('f', 168, long)
377 #define LL_IOC_LLOOP_ATTACH _IOWR('f', 169, long)
378 #define LL_IOC_LLOOP_DETACH _IOWR('f', 170, long)
379 #define LL_IOC_LLOOP_INFO _IOWR('f', 171, struct lu_fid)
380 #define LL_IOC_LLOOP_DETACH_BYDEV _IOWR('f', 172, long)
381 #define LL_IOC_PATH2FID _IOR ('f', 173, long)
382 #define LL_IOC_GET_CONNECT_FLAGS _IOWR('f', 174, __u64 *)
383 #define LL_IOC_GET_MDTIDX _IOR ('f', 175, int)
384 #define LL_IOC_FUTIMES_3 _IOWR('f', 176, struct ll_futimes_3)
385 #define LL_IOC_FLR_SET_MIRROR _IOW ('f', 177, long)
386 /* lustre_ioctl.h 177-210 */
387 #define LL_IOC_HSM_STATE_GET _IOR('f', 211, struct hsm_user_state)
388 #define LL_IOC_HSM_STATE_SET _IOW('f', 212, struct hsm_state_set)
389 #define LL_IOC_HSM_CT_START _IOW('f', 213, struct lustre_kernelcomm)
390 #define LL_IOC_HSM_COPY_START _IOW('f', 214, struct hsm_copy *)
391 #define LL_IOC_HSM_COPY_END _IOW('f', 215, struct hsm_copy *)
392 #define LL_IOC_HSM_PROGRESS _IOW('f', 216, struct hsm_user_request)
393 #define LL_IOC_HSM_REQUEST _IOW('f', 217, struct hsm_user_request)
394 #define LL_IOC_DATA_VERSION _IOR('f', 218, struct ioc_data_version)
395 #define LL_IOC_LOV_SWAP_LAYOUTS _IOW('f', 219, \
396 struct lustre_swap_layouts)
397 #define LL_IOC_HSM_ACTION _IOR('f', 220, \
398 struct hsm_current_action)
399 /* lustre_ioctl.h 221-232 */
400 #define LL_IOC_LMV_SETSTRIPE _IOWR('f', 240, struct lmv_user_md)
401 #define LL_IOC_LMV_GETSTRIPE _IOWR('f', 241, struct lmv_user_md)
402 #define LL_IOC_REMOVE_ENTRY _IOWR('f', 242, __u64)
403 #define LL_IOC_SET_LEASE _IOWR('f', 243, struct ll_ioc_lease)
404 #define LL_IOC_SET_LEASE_OLD _IOWR('f', 243, long)
405 #define LL_IOC_GET_LEASE _IO('f', 244)
406 #define LL_IOC_HSM_IMPORT _IOWR('f', 245, struct hsm_user_import)
407 #define LL_IOC_LMV_SET_DEFAULT_STRIPE _IOWR('f', 246, struct lmv_user_md)
408 #define LL_IOC_MIGRATE _IOR('f', 247, int)
409 #define LL_IOC_FID2MDTIDX _IOWR('f', 248, struct lu_fid)
410 #define LL_IOC_GETPARENT _IOWR('f', 249, struct getparent)
411 #define LL_IOC_LADVISE _IOR('f', 250, struct llapi_lu_ladvise)
413 #ifndef FS_IOC_FSGETXATTR
415 * Structure for FS_IOC_FSGETXATTR and FS_IOC_FSSETXATTR.
418 __u32 fsx_xflags; /* xflags field value (get/set) */
419 __u32 fsx_extsize; /* extsize field value (get/set)*/
420 __u32 fsx_nextents; /* nextents field value (get) */
421 __u32 fsx_projid; /* project identifier (get/set) */
422 unsigned char fsx_pad[12];
424 #define FS_IOC_FSGETXATTR _IOR('X', 31, struct fsxattr)
425 #define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
427 #define LL_IOC_FSGETXATTR FS_IOC_FSGETXATTR
428 #define LL_IOC_FSSETXATTR FS_IOC_FSSETXATTR
431 #define LL_STATFS_LMV 1
432 #define LL_STATFS_LOV 2
433 #define LL_STATFS_NODELAY 4
435 #define IOC_MDC_TYPE 'i'
436 #define IOC_MDC_LOOKUP _IOWR(IOC_MDC_TYPE, 20, struct obd_device *)
437 #define IOC_MDC_GETFILESTRIPE _IOWR(IOC_MDC_TYPE, 21, struct lov_user_md *)
438 #define IOC_MDC_GETFILEINFO _IOWR(IOC_MDC_TYPE, 22, struct lov_user_mds_data *)
439 #define LL_IOC_MDC_GETINFO _IOWR(IOC_MDC_TYPE, 23, struct lov_user_mds_data *)
441 #define MAX_OBD_NAME 128 /* If this changes, a NEW ioctl must be added */
443 /* Define O_LOV_DELAY_CREATE to be a mask that is not useful for regular
444 * files, but are unlikely to be used in practice and are not harmful if
445 * used incorrectly. O_NOCTTY and FASYNC are only meaningful for character
446 * devices and are safe for use on new files. See LU-4209. */
447 /* To be compatible with old statically linked binary we keep the check for
448 * the older 0100000000 flag. This is already removed upstream. LU-812. */
449 #define O_LOV_DELAY_CREATE_1_8 0100000000 /* FMODE_NONOTIFY masked in 2.6.36 */
450 #define O_LOV_DELAY_CREATE_MASK (O_NOCTTY | FASYNC)
451 #define O_LOV_DELAY_CREATE (O_LOV_DELAY_CREATE_1_8 | \
452 O_LOV_DELAY_CREATE_MASK)
454 #define LL_FILE_IGNORE_LOCK 0x00000001
455 #define LL_FILE_GROUP_LOCKED 0x00000002
456 #define LL_FILE_READAHEA 0x00000004
457 #define LL_FILE_LOCKED_DIRECTIO 0x00000008 /* client-side locks with dio */
458 #define LL_FILE_LOCKLESS_IO 0x00000010 /* server-side locks with cio */
460 #define LOV_USER_MAGIC_V1 0x0BD10BD0
461 #define LOV_USER_MAGIC LOV_USER_MAGIC_V1
462 #define LOV_USER_MAGIC_JOIN_V1 0x0BD20BD0
463 #define LOV_USER_MAGIC_V3 0x0BD30BD0
464 /* 0x0BD40BD0 is occupied by LOV_MAGIC_MIGRATE */
465 #define LOV_USER_MAGIC_SPECIFIC 0x0BD50BD0 /* for specific OSTs */
466 #define LOV_USER_MAGIC_COMP_V1 0x0BD60BD0
468 #define LMV_USER_MAGIC 0x0CD30CD0 /* default lmv magic */
469 #define LMV_USER_MAGIC_V0 0x0CD20CD0 /* old default lmv magic*/
471 #define LOV_PATTERN_NONE 0x000
472 #define LOV_PATTERN_RAID0 0x001
473 #define LOV_PATTERN_RAID1 0x002
474 #define LOV_PATTERN_MDT 0x100
475 #define LOV_PATTERN_CMOBD 0x200
477 #define LOV_PATTERN_F_MASK 0xffff0000
478 #define LOV_PATTERN_F_HOLE 0x40000000 /* there is hole in LOV EA */
479 #define LOV_PATTERN_F_RELEASED 0x80000000 /* HSM released file */
480 #define LOV_PATTERN_DEFAULT 0xffffffff
482 static inline bool lov_pattern_supported(__u32 pattern)
484 return pattern == LOV_PATTERN_RAID0 ||
485 pattern == LOV_PATTERN_MDT ||
486 pattern == (LOV_PATTERN_RAID0 | LOV_PATTERN_F_RELEASED);
489 #define LOV_MAXPOOLNAME 15
490 #define LOV_POOLNAMEF "%.15s"
492 #define LOV_MIN_STRIPE_BITS 16 /* maximum PAGE_SIZE (ia64), power of 2 */
493 #define LOV_MIN_STRIPE_SIZE (1 << LOV_MIN_STRIPE_BITS)
494 #define LOV_MAX_STRIPE_COUNT_OLD 160
495 /* This calculation is crafted so that input of 4096 will result in 160
496 * which in turn is equal to old maximal stripe count.
497 * XXX: In fact this is too simpified for now, what it also need is to get
498 * ea_type argument to clearly know how much space each stripe consumes.
500 * The limit of 12 pages is somewhat arbitrary, but is a reasonably large
501 * allocation that is sufficient for the current generation of systems.
503 * (max buffer size - lov+rpc header) / sizeof(struct lov_ost_data_v1) */
504 #define LOV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
505 #define LOV_ALL_STRIPES 0xffff /* only valid for directories */
506 #define LOV_V1_INSANE_STRIPE_COUNT 65532 /* maximum stripe count bz13933 */
508 #define XATTR_LUSTRE_PREFIX "lustre."
509 #define XATTR_LUSTRE_LOV XATTR_LUSTRE_PREFIX"lov"
511 #define lov_user_ost_data lov_user_ost_data_v1
512 struct lov_user_ost_data_v1 { /* per-stripe data structure */
513 struct ost_id l_ost_oi; /* OST object ID */
514 __u32 l_ost_gen; /* generation of this OST index */
515 __u32 l_ost_idx; /* OST index in LOV */
516 } __attribute__((packed));
518 #define lov_user_md lov_user_md_v1
519 struct lov_user_md_v1 { /* LOV EA user data (host-endian) */
520 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V1 */
521 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
522 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
523 __u32 lmm_stripe_size; /* size of stripe in bytes */
524 __u16 lmm_stripe_count; /* num stripes in use for this object */
526 __u16 lmm_stripe_offset; /* starting stripe offset in
527 * lmm_objects, use when writing */
528 __u16 lmm_layout_gen; /* layout generation number
529 * used when reading */
531 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
532 } __attribute__((packed, __may_alias__));
534 struct lov_user_md_v3 { /* LOV EA user data (host-endian) */
535 __u32 lmm_magic; /* magic number = LOV_USER_MAGIC_V3 */
536 __u32 lmm_pattern; /* LOV_PATTERN_RAID0, LOV_PATTERN_RAID1 */
537 struct ost_id lmm_oi; /* MDT parent inode id/seq (id/0 for 1.x) */
538 __u32 lmm_stripe_size; /* size of stripe in bytes */
539 __u16 lmm_stripe_count; /* num stripes in use for this object */
541 __u16 lmm_stripe_offset; /* starting stripe offset in
542 * lmm_objects, use when writing */
543 __u16 lmm_layout_gen; /* layout generation number
544 * used when reading */
546 char lmm_pool_name[LOV_MAXPOOLNAME + 1]; /* pool name */
547 struct lov_user_ost_data_v1 lmm_objects[0]; /* per-stripe data */
548 } __attribute__((packed));
555 #define DEXT "[%#llx, %#llx)"
556 #define PEXT(ext) (ext)->e_start, (ext)->e_end
558 static inline bool lu_extent_is_overlapped(struct lu_extent *e1,
559 struct lu_extent *e2)
561 return e1->e_start < e2->e_end && e2->e_start < e1->e_end;
564 static inline bool lu_extent_is_whole(struct lu_extent *e)
566 return e->e_start == 0 && e->e_end == LUSTRE_EOF;
569 enum lov_comp_md_entry_flags {
570 LCME_FL_PRIMARY = 0x00000001, /* Not used */
571 LCME_FL_STALE = 0x00000002, /* Not used */
572 LCME_FL_OFFLINE = 0x00000004, /* Not used */
573 LCME_FL_PREFERRED = 0x00000008, /* Not used */
574 LCME_FL_INIT = 0x00000010, /* instantiated */
575 LCME_FL_NEG = 0x80000000 /* used to indicate a negative flag,
576 won't be stored on disk */
579 #define LCME_KNOWN_FLAGS (LCME_FL_NEG | LCME_FL_INIT)
581 /* the highest bit in obdo::o_layout_version is used to mark if the file is
583 #define LU_LAYOUT_RESYNC LCME_FL_NEG
585 /* lcme_id can be specified as certain flags, and the the first
586 * bit of lcme_id is used to indicate that the ID is representing
587 * certain LCME_FL_* but not a real ID. Which implies we can have
588 * at most 31 flags (see LCME_FL_XXX). */
591 LCME_ID_MAX = 0x7FFFFFFF,
592 LCME_ID_ALL = 0xFFFFFFFF,
593 LCME_ID_NOT_ID = LCME_FL_NEG
596 #define LCME_ID_MASK LCME_ID_MAX
598 struct lov_comp_md_entry_v1 {
599 __u32 lcme_id; /* unique id of component */
600 __u32 lcme_flags; /* LCME_FL_XXX */
601 struct lu_extent lcme_extent; /* file extent for component */
602 __u32 lcme_offset; /* offset of component blob,
603 start from lov_comp_md_v1 */
604 __u32 lcme_size; /* size of component blob */
605 __u64 lcme_padding[2];
606 } __attribute__((packed));
608 #define SEQ_ID_MAX 0x0000FFFF
609 #define SEQ_ID_MASK SEQ_ID_MAX
610 /* bit 30:16 of lcme_id is used to store mirror id */
611 #define MIRROR_ID_MASK 0x7FFF0000
612 #define MIRROR_ID_SHIFT 16
614 static inline __u32 pflr_id(__u16 mirror_id, __u16 seqid)
616 return ((mirror_id << MIRROR_ID_SHIFT) & MIRROR_ID_MASK) | seqid;
619 static inline __u16 mirror_id_of(__u32 id)
621 return (id & MIRROR_ID_MASK) >> MIRROR_ID_SHIFT;
625 * on-disk data for lcm_flags. Valid if lcm_magic is LOV_MAGIC_COMP_V1.
627 enum lov_comp_md_flags {
628 /* the least 2 bits are used by FLR to record file state */
631 LCM_FL_WRITE_PENDING = 2,
632 LCM_FL_SYNC_PENDING = 3,
633 LCM_FL_FLR_MASK = 0x3,
636 struct lov_comp_md_v1 {
637 __u32 lcm_magic; /* LOV_USER_MAGIC_COMP_V1 */
638 __u32 lcm_size; /* overall size including this struct */
639 __u32 lcm_layout_gen;
641 __u16 lcm_entry_count;
642 /* lcm_mirror_count stores the number of actual mirrors minus 1,
643 * so that non-flr files will have value 0 meaning 1 mirror. */
644 __u16 lcm_mirror_count;
645 __u16 lcm_padding1[3];
647 struct lov_comp_md_entry_v1 lcm_entries[0];
648 } __attribute__((packed));
651 * Maximum number of mirrors Lustre can support.
653 #define LUSTRE_MIRROR_COUNT_MAX 16
655 static inline __u32 lov_user_md_size(__u16 stripes, __u32 lmm_magic)
657 if (stripes == (__u16)-1)
660 if (lmm_magic == LOV_USER_MAGIC_V1)
661 return sizeof(struct lov_user_md_v1) +
662 stripes * sizeof(struct lov_user_ost_data_v1);
663 return sizeof(struct lov_user_md_v3) +
664 stripes * sizeof(struct lov_user_ost_data_v1);
667 /* Compile with -D_LARGEFILE64_SOURCE or -D_GNU_SOURCE (or #define) to
668 * use this. It is unsafe to #define those values in this header as it
669 * is possible the application has already #included <sys/stat.h>. */
670 #ifdef HAVE_LOV_USER_MDS_DATA
671 #define lov_user_mds_data lov_user_mds_data_v1
672 struct lov_user_mds_data_v1 {
673 lstat_t lmd_st; /* MDS stat struct */
674 struct lov_user_md_v1 lmd_lmm; /* LOV EA V1 user data */
675 } __attribute__((packed));
677 struct lov_user_mds_data_v3 {
678 lstat_t lmd_st; /* MDS stat struct */
679 struct lov_user_md_v3 lmd_lmm; /* LOV EA V3 user data */
680 } __attribute__((packed));
683 struct lmv_user_mds_data {
684 struct lu_fid lum_fid;
690 LMV_HASH_TYPE_UNKNOWN = 0, /* 0 is reserved for testing purpose */
691 LMV_HASH_TYPE_ALL_CHARS = 1,
692 LMV_HASH_TYPE_FNV_1A_64 = 2,
696 #define LMV_HASH_NAME_ALL_CHARS "all_char"
697 #define LMV_HASH_NAME_FNV_1A_64 "fnv_1a_64"
699 extern char *mdt_hash_name[LMV_HASH_TYPE_MAX];
701 /* Got this according to how get LOV_MAX_STRIPE_COUNT, see above,
702 * (max buffer size - lmv+rpc header) / sizeof(struct lmv_user_mds_data) */
703 #define LMV_MAX_STRIPE_COUNT 2000 /* ((12 * 4096 - 256) / 24) */
704 #define lmv_user_md lmv_user_md_v1
705 struct lmv_user_md_v1 {
706 __u32 lum_magic; /* must be the first field */
707 __u32 lum_stripe_count; /* dirstripe count */
708 __u32 lum_stripe_offset; /* MDT idx for default dirstripe */
709 __u32 lum_hash_type; /* Dir stripe policy */
710 __u32 lum_type; /* LMV type: default or normal */
714 char lum_pool_name[LOV_MAXPOOLNAME + 1];
715 struct lmv_user_mds_data lum_objects[0];
716 } __attribute__((packed));
718 static inline int lmv_user_md_size(int stripes, int lmm_magic)
720 return sizeof(struct lmv_user_md) +
721 stripes * sizeof(struct lmv_user_mds_data);
724 struct ll_recreate_obj {
730 __u64 id; /* holds object id */
731 __u32 generation; /* holds object generation */
732 __u32 f_type; /* holds object type or stripe idx when passing it to
733 * OST for saving into EA. */
741 static inline bool obd_uuid_equals(const struct obd_uuid *u1,
742 const struct obd_uuid *u2)
744 return strcmp((char *)u1->uuid, (char *)u2->uuid) == 0;
747 static inline int obd_uuid_empty(struct obd_uuid *uuid)
749 return uuid->uuid[0] == '\0';
752 static inline void obd_str2uuid(struct obd_uuid *uuid, const char *tmp)
754 strncpy((char *)uuid->uuid, tmp, sizeof(*uuid));
755 uuid->uuid[sizeof(*uuid) - 1] = '\0';
758 /* For printf's only, make sure uuid is terminated */
759 static inline char *obd_uuid2str(const struct obd_uuid *uuid)
764 if (uuid->uuid[sizeof(*uuid) - 1] != '\0') {
765 /* Obviously not safe, but for printfs, no real harm done...
766 we're always null-terminated, even in a race. */
767 static char temp[sizeof(*uuid)];
768 memcpy(temp, uuid->uuid, sizeof(*uuid) - 1);
769 temp[sizeof(*uuid) - 1] = '\0';
772 return (char *)(uuid->uuid);
775 #define LUSTRE_MAXFSNAME 8
777 /* Extract fsname from uuid (or target name) of a target
778 e.g. (myfs-OST0007_UUID -> myfs)
779 see also deuuidify. */
780 static inline void obd_uuid2fsname(char *buf, char *uuid, int buflen)
784 strncpy(buf, uuid, buflen - 1);
785 buf[buflen - 1] = '\0';
786 p = strrchr(buf, '-');
791 /* printf display format for Lustre FIDs
792 * usage: printf("file FID is "DFID"\n", PFID(fid)); */
793 #define FID_NOBRACE_LEN 40
794 #define FID_LEN (FID_NOBRACE_LEN + 2)
795 #define DFID_NOBRACE "%#llx:0x%x:0x%x"
796 #define DFID "["DFID_NOBRACE"]"
797 #define PFID(fid) (unsigned long long)(fid)->f_seq, (fid)->f_oid, (fid)->f_ver
799 /* scanf input parse format for fids in DFID_NOBRACE format
800 * Need to strip '[' from DFID format first or use "["SFID"]" at caller.
801 * usage: sscanf(fidstr, SFID, RFID(&fid)); */
802 #define SFID "0x%llx:0x%x:0x%x"
803 #define RFID(fid) &((fid)->f_seq), &((fid)->f_oid), &((fid)->f_ver)
805 /********* Quotas **********/
807 #define LUSTRE_QUOTABLOCK_BITS 10
808 #define LUSTRE_QUOTABLOCK_SIZE (1 << LUSTRE_QUOTABLOCK_BITS)
810 static inline __u64 lustre_stoqb(size_t space)
812 return (space + LUSTRE_QUOTABLOCK_SIZE - 1) >> LUSTRE_QUOTABLOCK_BITS;
815 #define Q_QUOTACHECK 0x800100 /* deprecated as of 2.4 */
816 #define Q_INITQUOTA 0x800101 /* deprecated as of 2.4 */
817 #define Q_GETOINFO 0x800102 /* get obd quota info */
818 #define Q_GETOQUOTA 0x800103 /* get obd quotas */
819 #define Q_FINVALIDATE 0x800104 /* deprecated as of 2.4 */
821 /* these must be explicitly translated into linux Q_* in ll_dir_ioctl */
822 #define LUSTRE_Q_QUOTAON 0x800002 /* deprecated as of 2.4 */
823 #define LUSTRE_Q_QUOTAOFF 0x800003 /* deprecated as of 2.4 */
824 #define LUSTRE_Q_GETINFO 0x800005 /* get information about quota files */
825 #define LUSTRE_Q_SETINFO 0x800006 /* set information about quota files */
826 #define LUSTRE_Q_GETQUOTA 0x800007 /* get user quota structure */
827 #define LUSTRE_Q_SETQUOTA 0x800008 /* set user quota structure */
828 /* lustre-specific control commands */
829 #define LUSTRE_Q_INVALIDATE 0x80000b /* deprecated as of 2.4 */
830 #define LUSTRE_Q_FINVALIDATE 0x80000c /* deprecated as of 2.4 */
832 #define ALLQUOTA 255 /* set all quota */
833 static inline char *qtype_name(int qtype)
846 #define IDENTITY_DOWNCALL_MAGIC 0x6d6dd629
849 #define N_PERMS_MAX 64
851 struct perm_downcall_data {
857 struct identity_downcall_data {
864 struct perm_downcall_data idd_perms[N_PERMS_MAX];
868 #ifdef NEED_QUOTA_DEFS
870 #define QIF_BLIMITS 1
872 #define QIF_ILIMITS 4
876 #define QIF_LIMITS (QIF_BLIMITS | QIF_ILIMITS)
877 #define QIF_USAGE (QIF_SPACE | QIF_INODES)
878 #define QIF_TIMES (QIF_BTIME | QIF_ITIME)
879 #define QIF_ALL (QIF_LIMITS | QIF_USAGE | QIF_TIMES)
882 #endif /* !__KERNEL__ */
884 /* lustre volatile file support
885 * file name header: .^L^S^T^R:volatile"
887 #define LUSTRE_VOLATILE_HDR ".\x0c\x13\x14\x12:VOLATILE"
888 #define LUSTRE_VOLATILE_HDR_LEN 14
890 typedef enum lustre_quota_version {
892 } lustre_quota_version_t;
894 /* XXX: same as if_dqinfo struct in kernel */
902 /* XXX: same as if_dqblk struct in kernel, plus one padding */
904 __u64 dqb_bhardlimit;
905 __u64 dqb_bsoftlimit;
907 __u64 dqb_ihardlimit;
908 __u64 dqb_isoftlimit;
930 struct obd_dqinfo qc_dqinfo;
931 struct obd_dqblk qc_dqblk;
933 struct obd_uuid obd_uuid;
936 /* swap layout flags */
937 #define SWAP_LAYOUTS_CHECK_DV1 (1 << 0)
938 #define SWAP_LAYOUTS_CHECK_DV2 (1 << 1)
939 #define SWAP_LAYOUTS_KEEP_MTIME (1 << 2)
940 #define SWAP_LAYOUTS_KEEP_ATIME (1 << 3)
941 #define SWAP_LAYOUTS_CLOSE (1 << 4)
942 #define MERGE_LAYOUTS_CLOSE (1 << 5)
943 #define INTENT_LAYOUTS_CLOSE (SWAP_LAYOUTS_CLOSE | MERGE_LAYOUTS_CLOSE)
945 /* Swap XATTR_NAME_HSM as well, only on the MDT so far */
946 #define SWAP_LAYOUTS_MDS_HSM (1 << 31)
947 struct lustre_swap_layouts {
956 /********* Changelogs **********/
957 /** Changelog record types */
958 enum changelog_rec_type {
960 CL_CREATE = 1, /* namespace */
961 CL_MKDIR = 2, /* namespace */
962 CL_HARDLINK = 3, /* namespace */
963 CL_SOFTLINK = 4, /* namespace */
964 CL_MKNOD = 5, /* namespace */
965 CL_UNLINK = 6, /* namespace */
966 CL_RMDIR = 7, /* namespace */
967 CL_RENAME = 8, /* namespace */
968 CL_EXT = 9, /* namespace extended record (2nd half of rename) */
969 CL_OPEN = 10, /* not currently used */
970 CL_CLOSE = 11, /* may be written to log only with mtime change */
971 CL_LAYOUT = 12, /* file layout/striping modified */
975 CL_HSM = 16, /* HSM specific events, see flags */
976 CL_MTIME = 17, /* Precedence: setattr > mtime > ctime > atime */
980 CL_FLRW = 21, /* FLR: file was firstly written */
981 CL_RESYNC = 22, /* FLR: file was resync-ed */
985 static inline const char *changelog_type2str(int type) {
986 static const char *changelog_str[] = {
987 "MARK", "CREAT", "MKDIR", "HLINK", "SLINK", "MKNOD", "UNLNK",
988 "RMDIR", "RENME", "RNMTO", "OPEN", "CLOSE", "LYOUT", "TRUNC",
989 "SATTR", "XATTR", "HSM", "MTIME", "CTIME", "ATIME", "MIGRT",
993 if (type >= 0 && type < CL_LAST)
994 return changelog_str[type];
998 /* per-record flags */
999 #define CLF_FLAGSHIFT 12
1000 #define CLF_FLAGMASK ((1U << CLF_FLAGSHIFT) - 1)
1001 #define CLF_VERMASK (~CLF_FLAGMASK)
1002 enum changelog_rec_flags {
1003 CLF_VERSION = 0x1000,
1004 CLF_RENAME = 0x2000,
1006 CLF_EXTRA_FLAGS = 0x8000,
1007 CLF_SUPPORTED = CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS
1011 /* Anything under the flagmask may be per-type (if desired) */
1012 /* Flags for unlink */
1013 #define CLF_UNLINK_LAST 0x0001 /* Unlink of last hardlink */
1014 #define CLF_UNLINK_HSM_EXISTS 0x0002 /* File has something in HSM */
1015 /* HSM cleaning needed */
1016 /* Flags for rename */
1017 #define CLF_RENAME_LAST 0x0001 /* rename unlink last hardlink
1019 #define CLF_RENAME_LAST_EXISTS 0x0002 /* rename unlink last hardlink of target
1020 * has an archive in backend */
1023 /* 12b used (from high weight to low weight):
1028 #define CLF_HSM_ERR_L 0 /* HSM return code, 7 bits */
1029 #define CLF_HSM_ERR_H 6
1030 #define CLF_HSM_EVENT_L 7 /* HSM event, 3 bits, see enum hsm_event */
1031 #define CLF_HSM_EVENT_H 9
1032 #define CLF_HSM_FLAG_L 10 /* HSM flags, 2 bits, 1 used, 1 spare */
1033 #define CLF_HSM_FLAG_H 11
1034 #define CLF_HSM_SPARE_L 12 /* 4 spare bits */
1035 #define CLF_HSM_SPARE_H 15
1036 #define CLF_HSM_LAST 15
1038 /* Remove bits higher than _h, then extract the value
1039 * between _h and _l by shifting lower weigth to bit 0. */
1040 #define CLF_GET_BITS(_b, _h, _l) (((_b << (CLF_HSM_LAST - _h)) & 0xFFFF) \
1041 >> (CLF_HSM_LAST - _h + _l))
1043 #define CLF_HSM_SUCCESS 0x00
1044 #define CLF_HSM_MAXERROR 0x7E
1045 #define CLF_HSM_ERROVERFLOW 0x7F
1047 #define CLF_HSM_DIRTY 1 /* file is dirty after HSM request end */
1049 /* 3 bits field => 8 values allowed */
1061 static inline enum hsm_event hsm_get_cl_event(__u16 flags)
1063 return (enum hsm_event)CLF_GET_BITS(flags, CLF_HSM_EVENT_H,
1067 static inline void hsm_set_cl_event(int *flags, enum hsm_event he)
1069 *flags |= (he << CLF_HSM_EVENT_L);
1072 static inline __u16 hsm_get_cl_flags(int flags)
1074 return CLF_GET_BITS(flags, CLF_HSM_FLAG_H, CLF_HSM_FLAG_L);
1077 static inline void hsm_set_cl_flags(int *flags, int bits)
1079 *flags |= (bits << CLF_HSM_FLAG_L);
1082 static inline int hsm_get_cl_error(int flags)
1084 return CLF_GET_BITS(flags, CLF_HSM_ERR_H, CLF_HSM_ERR_L);
1087 static inline void hsm_set_cl_error(int *flags, int error)
1089 *flags |= (error << CLF_HSM_ERR_L);
1092 enum changelog_rec_extra_flags {
1094 CLFE_UIDGID = 0x0001,
1095 CLFE_SUPPORTED = CLFE_UIDGID
1098 enum changelog_send_flag {
1099 /* Not yet implemented */
1100 CHANGELOG_FLAG_FOLLOW = 0x01,
1101 /* Blocking IO makes sense in case of slow user parsing of the records,
1102 * but it also prevents us from cleaning up if the records are not
1104 CHANGELOG_FLAG_BLOCK = 0x02,
1105 /* Pack jobid into the changelog records if available. */
1106 CHANGELOG_FLAG_JOBID = 0x04,
1107 /* Pack additional flag bits into the changelog record */
1108 CHANGELOG_FLAG_EXTRA_FLAGS = 0x08,
1111 enum changelog_send_extra_flag {
1112 /* Pack uid/gid into the changelog record */
1113 CHANGELOG_EXTRA_FLAG_UIDGID = 0x01,
1116 #define CR_MAXSIZE cfs_size_round(2 * NAME_MAX + 2 + \
1117 changelog_rec_offset(CLF_SUPPORTED, \
1120 /* 31 usable bytes string + null terminator. */
1121 #define LUSTRE_JOBID_SIZE 32
1123 /* This is the minimal changelog record. It can contain extensions
1124 * such as rename fields or process jobid. Its exact content is described
1125 * by the cr_flags and cr_extra_flags.
1127 * Extensions are packed in the same order as their corresponding flags,
1128 * then in the same order as their corresponding extra flags.
1130 struct changelog_rec {
1132 __u16 cr_flags; /**< \a changelog_rec_flags */
1133 __u32 cr_type; /**< \a changelog_rec_type */
1134 __u64 cr_index; /**< changelog record number */
1135 __u64 cr_prev; /**< last index for this target fid */
1138 struct lu_fid cr_tfid; /**< target fid */
1139 __u32 cr_markerflags; /**< CL_MARK flags */
1141 struct lu_fid cr_pfid; /**< parent fid */
1144 /* Changelog extension for RENAME. */
1145 struct changelog_ext_rename {
1146 struct lu_fid cr_sfid; /**< source fid, or zero */
1147 struct lu_fid cr_spfid; /**< source parent fid, or zero */
1150 /* Changelog extension to include JOBID. */
1151 struct changelog_ext_jobid {
1152 char cr_jobid[LUSTRE_JOBID_SIZE]; /**< zero-terminated string. */
1155 /* Changelog extension to include additional flags. */
1156 struct changelog_ext_extra_flags {
1157 __u64 cr_extra_flags; /* Additional CLFE_* flags */
1160 /* Changelog extra extension to include UID/GID. */
1161 struct changelog_ext_uidgid {
1166 static inline struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1167 const struct changelog_rec *rec);
1169 static inline size_t changelog_rec_offset(enum changelog_rec_flags crf,
1170 enum changelog_rec_extra_flags cref)
1172 size_t size = sizeof(struct changelog_rec);
1174 if (crf & CLF_RENAME)
1175 size += sizeof(struct changelog_ext_rename);
1177 if (crf & CLF_JOBID)
1178 size += sizeof(struct changelog_ext_jobid);
1180 if (crf & CLF_EXTRA_FLAGS) {
1181 size += sizeof(struct changelog_ext_extra_flags);
1182 if (cref & CLFE_UIDGID)
1183 size += sizeof(struct changelog_ext_uidgid);
1189 static inline size_t changelog_rec_size(const struct changelog_rec *rec)
1191 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1193 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1194 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1196 return changelog_rec_offset(rec->cr_flags, cref);
1199 static inline size_t changelog_rec_varsize(const struct changelog_rec *rec)
1201 return changelog_rec_size(rec) - sizeof(*rec) + rec->cr_namelen;
1205 struct changelog_ext_rename *changelog_rec_rename(const struct changelog_rec *rec)
1207 enum changelog_rec_flags crf = rec->cr_flags & CLF_VERSION;
1209 return (struct changelog_ext_rename *)((char *)rec +
1210 changelog_rec_offset(crf,
1214 /* The jobid follows the rename extension, if present */
1216 struct changelog_ext_jobid *changelog_rec_jobid(const struct changelog_rec *rec)
1218 enum changelog_rec_flags crf = rec->cr_flags &
1219 (CLF_VERSION | CLF_RENAME);
1221 return (struct changelog_ext_jobid *)((char *)rec +
1222 changelog_rec_offset(crf,
1226 /* The additional flags follow the rename and jobid extensions, if present */
1228 struct changelog_ext_extra_flags *changelog_rec_extra_flags(
1229 const struct changelog_rec *rec)
1231 enum changelog_rec_flags crf = rec->cr_flags &
1232 (CLF_VERSION | CLF_RENAME | CLF_JOBID);
1234 return (struct changelog_ext_extra_flags *)((char *)rec +
1235 changelog_rec_offset(crf,
1239 /* The uid/gid is the first extra extension */
1241 struct changelog_ext_uidgid *changelog_rec_uidgid(
1242 const struct changelog_rec *rec)
1244 enum changelog_rec_flags crf = rec->cr_flags &
1245 (CLF_VERSION | CLF_RENAME | CLF_JOBID | CLF_EXTRA_FLAGS);
1247 return (struct changelog_ext_uidgid *)((char *)rec +
1248 changelog_rec_offset(crf,
1252 /* The name follows the rename, jobid and extra flags extns, if present */
1253 static inline char *changelog_rec_name(const struct changelog_rec *rec)
1255 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1257 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1258 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1260 return (char *)rec + changelog_rec_offset(rec->cr_flags & CLF_SUPPORTED,
1261 cref & CLFE_SUPPORTED);
1264 static inline size_t changelog_rec_snamelen(const struct changelog_rec *rec)
1266 return rec->cr_namelen - strlen(changelog_rec_name(rec)) - 1;
1269 static inline char *changelog_rec_sname(const struct changelog_rec *rec)
1271 char *cr_name = changelog_rec_name(rec);
1273 return cr_name + strlen(cr_name) + 1;
1277 * Remap a record to the desired format as specified by the crf flags.
1278 * The record must be big enough to contain the final remapped version.
1279 * Superfluous extension fields are removed and missing ones are added
1280 * and zeroed. The flags of the record are updated accordingly.
1282 * The jobid and rename extensions can be added to a record, to match the
1283 * format an application expects, typically. In this case, the newly added
1284 * fields will be zeroed.
1285 * The Jobid field can be removed, to guarantee compatibility with older
1286 * clients that don't expect this field in the records they process.
1288 * The following assumptions are being made:
1289 * - CLF_RENAME will not be removed
1290 * - CLF_JOBID will not be added without CLF_RENAME being added too
1291 * - CLF_EXTRA_FLAGS will not be added without CLF_JOBID being added too
1293 * @param[in,out] rec The record to remap.
1294 * @param[in] crf_wanted Flags describing the desired extensions.
1295 * @param[in] cref_want Flags describing the desired extra extensions.
1297 static inline void changelog_remap_rec(struct changelog_rec *rec,
1298 enum changelog_rec_flags crf_wanted,
1299 enum changelog_rec_extra_flags cref_want)
1301 char *uidgid_mov = NULL;
1305 enum changelog_rec_extra_flags cref = CLFE_INVALID;
1307 crf_wanted &= CLF_SUPPORTED;
1308 cref_want &= CLFE_SUPPORTED;
1310 if ((rec->cr_flags & CLF_SUPPORTED) == crf_wanted) {
1311 if (!(rec->cr_flags & CLF_EXTRA_FLAGS) ||
1312 (rec->cr_flags & CLF_EXTRA_FLAGS &&
1313 (changelog_rec_extra_flags(rec)->cr_extra_flags &
1319 /* First move the variable-length name field */
1320 memmove((char *)rec + changelog_rec_offset(crf_wanted, cref_want),
1321 changelog_rec_name(rec), rec->cr_namelen);
1323 /* Locations of extensions in the remapped record */
1324 if (rec->cr_flags & CLF_EXTRA_FLAGS) {
1325 uidgid_mov = (char *)rec +
1326 changelog_rec_offset(crf_wanted & CLF_SUPPORTED,
1328 cref = changelog_rec_extra_flags(rec)->cr_extra_flags;
1331 ef_mov = (char *)rec +
1332 changelog_rec_offset(crf_wanted & ~CLF_EXTRA_FLAGS,
1334 jid_mov = (char *)rec +
1335 changelog_rec_offset(crf_wanted &
1336 ~(CLF_EXTRA_FLAGS | CLF_JOBID),
1338 rnm_mov = (char *)rec +
1339 changelog_rec_offset(crf_wanted &
1345 /* Move the extension fields to the desired positions */
1346 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1347 (rec->cr_flags & CLF_EXTRA_FLAGS)) {
1348 if ((cref_want & CLFE_UIDGID) && (cref & CLFE_UIDGID))
1349 memmove(uidgid_mov, changelog_rec_uidgid(rec),
1350 sizeof(struct changelog_ext_uidgid));
1352 memmove(ef_mov, changelog_rec_extra_flags(rec),
1353 sizeof(struct changelog_ext_extra_flags));
1356 if ((crf_wanted & CLF_JOBID) && (rec->cr_flags & CLF_JOBID))
1357 memmove(jid_mov, changelog_rec_jobid(rec),
1358 sizeof(struct changelog_ext_jobid));
1360 if ((crf_wanted & CLF_RENAME) && (rec->cr_flags & CLF_RENAME))
1361 memmove(rnm_mov, changelog_rec_rename(rec),
1362 sizeof(struct changelog_ext_rename));
1364 /* Clear newly added fields */
1365 if (uidgid_mov && (cref_want & CLFE_UIDGID) &&
1366 !(cref & CLFE_UIDGID))
1367 memset(uidgid_mov, 0, sizeof(struct changelog_ext_uidgid));
1369 if ((crf_wanted & CLF_EXTRA_FLAGS) &&
1370 !(rec->cr_flags & CLF_EXTRA_FLAGS))
1371 memset(ef_mov, 0, sizeof(struct changelog_ext_extra_flags));
1373 if ((crf_wanted & CLF_JOBID) && !(rec->cr_flags & CLF_JOBID))
1374 memset(jid_mov, 0, sizeof(struct changelog_ext_jobid));
1376 if ((crf_wanted & CLF_RENAME) && !(rec->cr_flags & CLF_RENAME))
1377 memset(rnm_mov, 0, sizeof(struct changelog_ext_rename));
1379 /* Update the record's flags accordingly */
1380 rec->cr_flags = (rec->cr_flags & CLF_FLAGMASK) | crf_wanted;
1381 if (rec->cr_flags & CLF_EXTRA_FLAGS)
1382 changelog_rec_extra_flags(rec)->cr_extra_flags =
1383 changelog_rec_extra_flags(rec)->cr_extra_flags |
1387 enum changelog_message_type {
1388 CL_RECORD = 10, /* message is a changelog_rec */
1389 CL_EOF = 11, /* at end of current changelog */
1392 /********* Misc **********/
1394 struct ioc_data_version {
1396 __u32 idv_layout_version; /* FLR: layout version for OST objects */
1397 __u32 idv_flags; /* enum ioc_data_version_flags */
1400 enum ioc_data_version_flags {
1401 LL_DV_RD_FLUSH = (1 << 0), /* Flush dirty pages from clients */
1402 LL_DV_WR_FLUSH = (1 << 1), /* Flush all caching pages from clients */
1406 #define offsetof(typ, memb) ((unsigned long)((char *)&(((typ *)0)->memb)))
1409 #define dot_lustre_name ".lustre"
1412 /********* HSM **********/
1414 /** HSM per-file state
1415 * See HSM_FLAGS below.
1418 HS_NONE = 0x00000000,
1419 HS_EXISTS = 0x00000001,
1420 HS_DIRTY = 0x00000002,
1421 HS_RELEASED = 0x00000004,
1422 HS_ARCHIVED = 0x00000008,
1423 HS_NORELEASE = 0x00000010,
1424 HS_NOARCHIVE = 0x00000020,
1425 HS_LOST = 0x00000040,
1428 /* HSM user-setable flags. */
1429 #define HSM_USER_MASK (HS_NORELEASE | HS_NOARCHIVE | HS_DIRTY)
1431 /* Other HSM flags. */
1432 #define HSM_STATUS_MASK (HS_EXISTS | HS_LOST | HS_RELEASED | HS_ARCHIVED)
1435 * All HSM-related possible flags that could be applied to a file.
1436 * This should be kept in sync with hsm_states.
1438 #define HSM_FLAGS_MASK (HSM_USER_MASK | HSM_STATUS_MASK)
1441 * HSMÂ request progress state
1443 enum hsm_progress_states {
1450 static inline const char *hsm_progress_state2name(enum hsm_progress_states s)
1453 case HPS_WAITING: return "waiting";
1454 case HPS_RUNNING: return "running";
1455 case HPS_DONE: return "done";
1456 default: return "unknown";
1463 } __attribute__((packed));
1466 * Current HSM states of a Lustre file.
1468 * This structure purpose is to be sent to user-space mainly. It describes the
1469 * current HSM flags and in-progress action.
1471 struct hsm_user_state {
1472 /** Current HSM states, from enum hsm_states. */
1474 __u32 hus_archive_id;
1475 /** The current undergoing action, if there is one */
1476 __u32 hus_in_progress_state;
1477 __u32 hus_in_progress_action;
1478 struct hsm_extent hus_in_progress_location;
1479 char hus_extended_info[];
1482 struct hsm_state_set_ioc {
1483 struct lu_fid hssi_fid;
1485 __u64 hssi_clearmask;
1489 * This structure describes the current in-progress action for a file.
1490 * it is retuned to user space and send over the wire
1492 struct hsm_current_action {
1493 /** The current undergoing action, if there is one */
1494 /* state is one of hsm_progress_states */
1496 /* action is one of hsm_user_action */
1498 struct hsm_extent hca_location;
1501 /***** HSM user requests ******/
1502 /* User-generated (lfs/ioctl) request types */
1503 enum hsm_user_action {
1504 HUA_NONE = 1, /* no action (noop) */
1505 HUA_ARCHIVE = 10, /* copy to hsm */
1506 HUA_RESTORE = 11, /* prestage */
1507 HUA_RELEASE = 12, /* drop ost objects */
1508 HUA_REMOVE = 13, /* remove from archive */
1509 HUA_CANCEL = 14 /* cancel a request */
1512 static inline const char *hsm_user_action2name(enum hsm_user_action a)
1515 case HUA_NONE: return "NOOP";
1516 case HUA_ARCHIVE: return "ARCHIVE";
1517 case HUA_RESTORE: return "RESTORE";
1518 case HUA_RELEASE: return "RELEASE";
1519 case HUA_REMOVE: return "REMOVE";
1520 case HUA_CANCEL: return "CANCEL";
1521 default: return "UNKNOWN";
1526 * List of hr_flags (bit field)
1528 #define HSM_FORCE_ACTION 0x0001
1529 /* used by CT, cannot be set by user */
1530 #define HSM_GHOST_COPY 0x0002
1533 * Contains all the fixed part of struct hsm_user_request.
1536 struct hsm_request {
1537 __u32 hr_action; /* enum hsm_user_action */
1538 __u32 hr_archive_id; /* archive id, used only with HUA_ARCHIVE */
1539 __u64 hr_flags; /* request flags */
1540 __u32 hr_itemcount; /* item count in hur_user_item vector */
1544 struct hsm_user_item {
1545 struct lu_fid hui_fid;
1546 struct hsm_extent hui_extent;
1547 } __attribute__((packed));
1549 struct hsm_user_request {
1550 struct hsm_request hur_request;
1551 struct hsm_user_item hur_user_item[0];
1552 /* extra data blob at end of struct (after all
1553 * hur_user_items), only use helpers to access it
1555 } __attribute__((packed));
1557 /** Return pointer to data field in a hsm user request */
1558 static inline void *hur_data(struct hsm_user_request *hur)
1560 return &(hur->hur_user_item[hur->hur_request.hr_itemcount]);
1564 * Compute the current length of the provided hsm_user_request. This returns -1
1565 * instead of an errno because ssize_t is defined to be only [ -1, SSIZE_MAX ]
1567 * return -1 on bounds check error.
1569 static inline ssize_t hur_len(struct hsm_user_request *hur)
1573 /* can't overflow a __u64 since hr_itemcount is only __u32 */
1574 size = offsetof(struct hsm_user_request, hur_user_item[0]) +
1575 (__u64)hur->hur_request.hr_itemcount *
1576 sizeof(hur->hur_user_item[0]) + hur->hur_request.hr_data_len;
1578 if (size != (ssize_t)size)
1584 /****** HSM RPCs to copytool *****/
1585 /* Message types the copytool may receive */
1586 enum hsm_message_type {
1587 HMT_ACTION_LIST = 100, /* message is a hsm_action_list */
1590 /* Actions the copytool may be instructed to take for a given action_item */
1591 enum hsm_copytool_action {
1592 HSMA_NONE = 10, /* no action */
1593 HSMA_ARCHIVE = 20, /* arbitrary offset */
1599 static inline const char *hsm_copytool_action2name(enum hsm_copytool_action a)
1602 case HSMA_NONE: return "NOOP";
1603 case HSMA_ARCHIVE: return "ARCHIVE";
1604 case HSMA_RESTORE: return "RESTORE";
1605 case HSMA_REMOVE: return "REMOVE";
1606 case HSMA_CANCEL: return "CANCEL";
1607 default: return "UNKNOWN";
1611 /* Copytool item action description */
1612 struct hsm_action_item {
1613 __u32 hai_len; /* valid size of this struct */
1614 __u32 hai_action; /* hsm_copytool_action, but use known size */
1615 struct lu_fid hai_fid; /* Lustre FID to operate on */
1616 struct lu_fid hai_dfid; /* fid used for data access */
1617 struct hsm_extent hai_extent; /* byte range to operate on */
1618 __u64 hai_cookie; /* action cookie from coordinator */
1619 __u64 hai_gid; /* grouplock id */
1620 char hai_data[0]; /* variable length */
1621 } __attribute__((packed));
1624 * helper function which print in hexa the first bytes of
1627 * \param hai [IN] record to print
1628 * \param buffer [IN,OUT] buffer to write the hex string to
1629 * \param len [IN] max buffer length
1633 static inline char *hai_dump_data_field(const struct hsm_action_item *hai,
1634 char *buffer, size_t len)
1641 data_len = hai->hai_len - sizeof(*hai);
1642 for (i = 0; (i < data_len) && (len > 2); i++) {
1643 snprintf(ptr, 3, "%02X", (unsigned char)hai->hai_data[i]);
1653 /* Copytool action list */
1654 #define HAL_VERSION 1
1655 #define HAL_MAXSIZE LNET_MTU /* bytes, used in userspace only */
1656 struct hsm_action_list {
1658 __u32 hal_count; /* number of hai's to follow */
1659 __u64 hal_compound_id; /* returned by coordinator */
1661 __u32 hal_archive_id; /* which archive backend */
1663 char hal_fsname[0]; /* null-terminated */
1664 /* struct hsm_action_item[hal_count] follows, aligned on 8-byte
1665 boundaries. See hai_zero */
1666 } __attribute__((packed));
1668 #ifndef HAVE_CFS_SIZE_ROUND
1669 static inline int cfs_size_round (int val)
1671 return (val + 7) & (~0x7);
1673 #define HAVE_CFS_SIZE_ROUND
1676 /* Return pointer to first hai in action list */
1677 static inline struct hsm_action_item *hai_first(struct hsm_action_list *hal)
1679 return (struct hsm_action_item *)(hal->hal_fsname +
1680 cfs_size_round(strlen(hal-> \
1684 /* Return pointer to next hai */
1685 static inline struct hsm_action_item * hai_next(struct hsm_action_item *hai)
1687 return (struct hsm_action_item *)((char *)hai +
1688 cfs_size_round(hai->hai_len));
1691 /* Return size of an hsm_action_list */
1692 static inline size_t hal_size(struct hsm_action_list *hal)
1696 struct hsm_action_item *hai;
1698 sz = sizeof(*hal) + cfs_size_round(strlen(hal->hal_fsname) + 1);
1699 hai = hai_first(hal);
1700 for (i = 0; i < hal->hal_count ; i++, hai = hai_next(hai))
1701 sz += cfs_size_round(hai->hai_len);
1707 * describe the attributes to be set on imported file
1709 struct hsm_user_import {
1718 __u32 hui_archive_id;
1721 /* Copytool progress reporting */
1722 #define HP_FLAG_COMPLETED 0x01
1723 #define HP_FLAG_RETRY 0x02
1725 struct hsm_progress {
1726 struct lu_fid hp_fid;
1728 struct hsm_extent hp_extent;
1730 __u16 hp_errval; /* positive val */
1735 __u64 hc_data_version;
1737 __u16 hc_errval; /* positive val */
1739 struct hsm_action_item hc_hai;
1743 enum llapi_json_types {
1744 LLAPI_JSON_INTEGER = 1,
1750 struct llapi_json_item {
1759 struct llapi_json_item *lji_next;
1762 struct llapi_json_item_list {
1763 int ljil_item_count;
1764 struct llapi_json_item *ljil_items;
1767 enum lu_ladvise_type {
1768 LU_LADVISE_INVALID = 0,
1769 LU_LADVISE_WILLREAD = 1,
1770 LU_LADVISE_DONTNEED = 2,
1771 LU_LADVISE_LOCKNOEXPAND = 3,
1772 LU_LADVISE_LOCKAHEAD = 4,
1776 #define LU_LADVISE_NAMES { \
1777 [LU_LADVISE_WILLREAD] = "willread", \
1778 [LU_LADVISE_DONTNEED] = "dontneed", \
1779 [LU_LADVISE_LOCKNOEXPAND] = "locknoexpand", \
1780 [LU_LADVISE_LOCKAHEAD] = "lockahead", \
1783 /* This is the userspace argument for ladvise. It is currently the same as
1784 * what goes on the wire (struct lu_ladvise), but is defined separately as we
1785 * may need info which is only used locally. */
1786 struct llapi_lu_ladvise {
1787 __u16 lla_advice; /* advice type */
1788 __u16 lla_value1; /* values for different advice types */
1790 __u64 lla_start; /* first byte of extent for advice */
1791 __u64 lla_end; /* last byte of extent for advice */
1797 LF_ASYNC = 0x00000001,
1798 LF_UNSET = 0x00000002,
1801 #define LADVISE_MAGIC 0x1ADF1CE0
1802 /* Masks of valid flags for each advice */
1803 #define LF_LOCKNOEXPAND_MASK LF_UNSET
1804 /* Flags valid for all advices not explicitly specified */
1805 #define LF_DEFAULT_MASK LF_ASYNC
1807 #define LF_MASK (LF_ASYNC | LF_UNSET)
1809 #define lla_lockahead_mode lla_value1
1810 #define lla_peradvice_flags lla_value2
1811 #define lla_lockahead_result lla_value3
1813 /* This is the userspace argument for ladvise, corresponds to ladvise_hdr which
1814 * is used on the wire. It is defined separately as we may need info which is
1815 * only used locally. */
1816 struct llapi_ladvise_hdr {
1817 __u32 lah_magic; /* LADVISE_MAGIC */
1818 __u32 lah_count; /* number of advices */
1819 __u64 lah_flags; /* from enum ladvise_flag */
1820 __u32 lah_value1; /* unused */
1821 __u32 lah_value2; /* unused */
1822 __u64 lah_value3; /* unused */
1823 struct llapi_lu_ladvise lah_advise[0]; /* advices in this header */
1826 #define LAH_COUNT_MAX (1024)
1830 SK_CRYPT_INVALID = -1,
1832 SK_CRYPT_AES256_CTR = 1,
1837 SK_HMAC_INVALID = -1,
1844 struct sk_crypt_type {
1849 struct sk_hmac_type {
1854 enum lock_mode_user {
1860 #define LOCK_MODE_NAMES { \
1861 [MODE_READ_USER] = "READ",\
1862 [MODE_WRITE_USER] = "WRITE"\
1865 enum lockahead_results {
1866 LLA_RESULT_SENT = 0,
1867 LLA_RESULT_DIFFERENT,
1871 #if defined(__cplusplus)
1875 /** @} lustreuser */
1877 #endif /* _LUSTRE_USER_H */